Final Project.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "b2fde6c3",
   "metadata": {},
   "source": [
    "**Review**\n",
    "\n",
    "Hi, my name is Dmitry and I will be reviewing your code.\n",
    "  \n",
    "You can find my comments in colored markdown cells:\n",
    "  \n",
    "<div class=\"alert alert-success\">\n",
    "  If everything is done successfully.\n",
    "</div>\n",
    "  \n",
    "<div class=\"alert alert-warning\">\n",
    "  If I have some (optional) suggestions, or questions to think about, or general comments.\n",
    "</div>\n",
    "  \n",
    "<div class=\"alert alert-danger\">\n",
    "  If a section requires some corrections. Work can't be accepted with red comments.\n",
    "</div>\n",
    "  \n",
    "Please don't remove my comments, as it will make further review iterations much harder for me.\n",
    "  \n",
    "Feel free to reply to my comments or ask questions using the following template:\n",
    "  \n",
    "<div class=\"alert alert-info\">\n",
    "  For your comments and questions.\n",
    "</div>\n",
    "  \n",
    "First of all, thank you for turning in the project! You did a great job overall! I added some suggestions for improvements below, but the code can be accepted now. Based on your final model's ROC-AUC you've got 6 SP. Congratulations! :)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a77ff76f",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\">\n",
    "    Hi Johannah! I have approved your initial report for the final project. Great work! Just wanted to add some helpful pointers:\n",
    "    <br/><br/>\n",
    "If end date is empty, assume that the customer is still with the company (no churn)\n",
    "    <br/><br/>\n",
    "    When merging dataframes make sure that you do not miss any customers since some ID’s may not exist in certain tables. You can treat these missing cases as that customer not signing up for the specific service\n",
    "<br/><br/>\n",
    "    Your target variable should be Churn / No Churn, this will be a binary classification task\n",
    "<br/><br/>\n",
    "    Make sure that the validation set is used to tune hyperparameters, good idea to use gridsearch\n",
    "<br/><br/>\n",
    "    There does indeed exist a class imbalance in the data, make sure this is accounted for\n",
    "<br/><br/>\n",
    "    Careful with data leakage, make sure to not include features that may be collinear (i.e if using one-hot encoding for categorical variables, drop one of the columns) or reference the target variable in some way (see topic on data leakage)\n",
    "<br/><br/>\n",
    "    Some models you can try are logistic regression, random forest, gradient boosting algorithms like catboost\n",
    "<br/><br/>\n",
    "    Very good EDA plan\n",
    "<br/><br/>\n",
    "    The goal of this project is to understand why customers churned or stopped their service. And we can use machine learning to identify features of the data that had high predictive power when determining which customers will churn. And if we know that, what can we do as a company to prevent ongoing churn (this part is more of a discussion question for your end\n",
    "<br/><br/>\n",
    "-Yervand, Data Science Tutor\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "153e5b36",
   "metadata": {},
   "source": [
    "# Telecom Churn Prediction\n",
    "The telecom operator Interconnect would like to be able to forecast their churn of clients. If it's discovered that a user is planning to leave, they will be offered promotional codes and special plan options. Interconnect's marketing team has collected some of their clientele's personal data, including information about their plans and contracts.\n",
    "\n",
    "### Interconnect's services\n",
    "\n",
    "Interconnect mainly provides two types of services:\n",
    "\n",
    "1. Landline communication. The telephone can be connected to several lines simultaneously.\n",
    "2. Internet. The network can be set up via a telephone line (DSL, *digital subscriber line*) or through a fiber optic cable.\n",
    "\n",
    "Some other services the company provides include:\n",
    "\n",
    "- Internet security: antivirus software (*DeviceProtection*) and a malicious website blocker (*OnlineSecurity*)\n",
    "- A dedicated technical support line (*TechSupport*)\n",
    "- Cloud file storage and data backup (*OnlineBackup*)\n",
    "- TV streaming (*StreamingTV*) and a movie directory (*StreamingMovies*)\n",
    "\n",
    "The clients can choose either a monthly payment or sign a 1- or 2-year contract. They can use various payment methods and receive an electronic invoice after a transaction.\n",
    "\n",
    "### Data Description\n",
    "\n",
    "The data consists of files obtained from different sources:\n",
    "\n",
    "- `contract.csv` — contract information\n",
    "- `personal.csv` — the client's personal data\n",
    "- `internet.csv` — information about Internet services\n",
    "- `phone.csv` — information about telephone services\n",
    "\n",
    "In each file, the column `customerID` contains a unique code assigned to each client."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "19c95581",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import time\n",
    "from datetime import datetime\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "import seaborn as sns\n",
    "\n",
    "from sklearn.model_selection import train_test_split, GridSearchCV\n",
    "\n",
    "\n",
    "from sklearn.preprocessing import OrdinalEncoder\n",
    "from sklearn.preprocessing import OneHotEncoder\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "pd.options.mode.chained_assignment = None # to avoid SettingWithCopyWarning after scaling\n",
    "\n",
    "\n",
    "from sklearn.dummy import DummyClassifier \n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from sklearn.linear_model import LogisticRegression \n",
    "from sklearn.ensemble import RandomForestClassifier \n",
    "from catboost import CatBoostClassifier \n",
    "from lightgbm import LGBMClassifier \n",
    "from xgboost import XGBClassifier \n",
    "\n",
    "\n",
    "from sklearn import metrics\n",
    "from sklearn.metrics import *\n",
    "from sklearn.metrics import accuracy_score\n",
    "from sklearn.metrics import confusion_matrix, classification_report\n",
    "from sklearn.metrics import precision_score, recall_score, f1_score, precision_recall_curve\n",
    "from sklearn.metrics import balanced_accuracy_score, roc_auc_score\n",
    "\n",
    "# import warnings\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "\n",
    "# import sklearn utilities\n",
    "from sklearn.utils import shuffle\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "0f1832b0",
   "metadata": {},
   "outputs": [],
   "source": [
    "contracts_df= pd.read_csv('/datasets/final_provider/contract.csv')\n",
    "personal_df = pd.read_csv('/datasets/final_provider/personal.csv')\n",
    "internet_df = pd.read_csv('/datasets/final_provider/internet.csv')\n",
    "phone_df = pd.read_csv('/datasets/final_provider/phone.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "6f46454d",
   "metadata": {},
   "outputs": [],
   "source": [
    "def draw_general_info(data):\n",
    "    \"\"\"Takes dataframe and draws the main staistics\n",
    "    and information about the data in one table.\n",
    "    \"\"\"\n",
    "    df = pd.DataFrame()\n",
    "    df[\"missing\"] = data.isna().sum().sort_values(ascending=False)\n",
    "    df[\"%_missing\"] = (data.isna().sum() / data.isna().count()).sort_values(\n",
    "        ascending=False\n",
    "    )\n",
    "    df[\"%_missing\"] = pd.Series(\n",
    "        [\"{:.3%}\".format(x) for x in df[\"%_missing\"]], index=df.index\n",
    "    )\n",
    "    df[\"value_sample\"] = data.sample().T\n",
    "    df[\"dtype\"] = data.dtypes\n",
    "    describe = data.describe(include=\"all\").round(3).T\n",
    "    df = df.merge(describe, left_on=df.index, right_on=describe.index)\n",
    "    df[\"count\"] = df[\"count\"].astype(int)\n",
    "    df.rename({\"key_0\":\"column_name\"}, inplace=True, axis=1)\n",
    "    df[\"unique\"] = pd.Series(\n",
    "        [data[column].nunique() for column in df[\"column_name\"]], index=df.index\n",
    "    )\n",
    "    df.fillna(\"\", inplace=True)\n",
    "    display(df)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "5a828810",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>column_name</th>\n",
       "      <th>missing</th>\n",
       "      <th>%_missing</th>\n",
       "      <th>value_sample</th>\n",
       "      <th>dtype</th>\n",
       "      <th>count</th>\n",
       "      <th>unique</th>\n",
       "      <th>top</th>\n",
       "      <th>freq</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>min</th>\n",
       "      <th>25%</th>\n",
       "      <th>50%</th>\n",
       "      <th>75%</th>\n",
       "      <th>max</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>customerID</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>5103-MHMHY</td>\n",
       "      <td>object</td>\n",
       "      <td>7043</td>\n",
       "      <td>7043</td>\n",
       "      <td>5996-DAOQL</td>\n",
       "      <td>1</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>BeginDate</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>2019-12-01</td>\n",
       "      <td>object</td>\n",
       "      <td>7043</td>\n",
       "      <td>77</td>\n",
       "      <td>2014-02-01</td>\n",
       "      <td>366</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>EndDate</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>2020-01-01 00:00:00</td>\n",
       "      <td>object</td>\n",
       "      <td>7043</td>\n",
       "      <td>5</td>\n",
       "      <td>No</td>\n",
       "      <td>5174</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Type</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>object</td>\n",
       "      <td>7043</td>\n",
       "      <td>3</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>3875</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>PaperlessBilling</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>No</td>\n",
       "      <td>object</td>\n",
       "      <td>7043</td>\n",
       "      <td>2</td>\n",
       "      <td>Yes</td>\n",
       "      <td>4171</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>PaymentMethod</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>Mailed check</td>\n",
       "      <td>object</td>\n",
       "      <td>7043</td>\n",
       "      <td>4</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>2365</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>MonthlyCharges</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>45.95</td>\n",
       "      <td>float64</td>\n",
       "      <td>7043</td>\n",
       "      <td>1585</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td>64.762</td>\n",
       "      <td>30.09</td>\n",
       "      <td>18.25</td>\n",
       "      <td>35.5</td>\n",
       "      <td>70.35</td>\n",
       "      <td>89.85</td>\n",
       "      <td>118.75</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>TotalCharges</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>45.95</td>\n",
       "      <td>object</td>\n",
       "      <td>7043</td>\n",
       "      <td>6531</td>\n",
       "      <td>20.2</td>\n",
       "      <td>11</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        column_name  missing %_missing         value_sample    dtype  count  \\\n",
       "0        customerID        0    0.000%           5103-MHMHY   object   7043   \n",
       "1         BeginDate        0    0.000%           2019-12-01   object   7043   \n",
       "2           EndDate        0    0.000%  2020-01-01 00:00:00   object   7043   \n",
       "3              Type        0    0.000%       Month-to-month   object   7043   \n",
       "4  PaperlessBilling        0    0.000%                   No   object   7043   \n",
       "5     PaymentMethod        0    0.000%         Mailed check   object   7043   \n",
       "6    MonthlyCharges        0    0.000%                45.95  float64   7043   \n",
       "7      TotalCharges        0    0.000%                45.95   object   7043   \n",
       "\n",
       "   unique               top  freq    mean    std    min   25%    50%    75%  \\\n",
       "0    7043        5996-DAOQL     1                                             \n",
       "1      77        2014-02-01   366                                             \n",
       "2       5                No  5174                                             \n",
       "3       3    Month-to-month  3875                                             \n",
       "4       2               Yes  4171                                             \n",
       "5       4  Electronic check  2365                                             \n",
       "6    1585                          64.762  30.09  18.25  35.5  70.35  89.85   \n",
       "7    6531              20.2    11                                             \n",
       "\n",
       "      max  \n",
       "0          \n",
       "1          \n",
       "2          \n",
       "3          \n",
       "4          \n",
       "5          \n",
       "6  118.75  \n",
       "7          "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "draw_general_info(contracts_df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "3221bc7b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>customerID</th>\n",
       "      <th>BeginDate</th>\n",
       "      <th>EndDate</th>\n",
       "      <th>Type</th>\n",
       "      <th>PaperlessBilling</th>\n",
       "      <th>PaymentMethod</th>\n",
       "      <th>MonthlyCharges</th>\n",
       "      <th>TotalCharges</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>7590-VHVEG</td>\n",
       "      <td>2020-01-01</td>\n",
       "      <td>No</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>29.85</td>\n",
       "      <td>29.85</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>5575-GNVDE</td>\n",
       "      <td>2017-04-01</td>\n",
       "      <td>No</td>\n",
       "      <td>One year</td>\n",
       "      <td>No</td>\n",
       "      <td>Mailed check</td>\n",
       "      <td>56.95</td>\n",
       "      <td>1889.5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3668-QPYBK</td>\n",
       "      <td>2019-10-01</td>\n",
       "      <td>2019-12-01 00:00:00</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Mailed check</td>\n",
       "      <td>53.85</td>\n",
       "      <td>108.15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>7795-CFOCW</td>\n",
       "      <td>2016-05-01</td>\n",
       "      <td>No</td>\n",
       "      <td>One year</td>\n",
       "      <td>No</td>\n",
       "      <td>Bank transfer (automatic)</td>\n",
       "      <td>42.30</td>\n",
       "      <td>1840.75</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>9237-HQITU</td>\n",
       "      <td>2019-09-01</td>\n",
       "      <td>2019-11-01 00:00:00</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>70.70</td>\n",
       "      <td>151.65</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   customerID   BeginDate              EndDate            Type  \\\n",
       "0  7590-VHVEG  2020-01-01                   No  Month-to-month   \n",
       "1  5575-GNVDE  2017-04-01                   No        One year   \n",
       "2  3668-QPYBK  2019-10-01  2019-12-01 00:00:00  Month-to-month   \n",
       "3  7795-CFOCW  2016-05-01                   No        One year   \n",
       "4  9237-HQITU  2019-09-01  2019-11-01 00:00:00  Month-to-month   \n",
       "\n",
       "  PaperlessBilling              PaymentMethod  MonthlyCharges TotalCharges  \n",
       "0              Yes           Electronic check           29.85        29.85  \n",
       "1               No               Mailed check           56.95       1889.5  \n",
       "2              Yes               Mailed check           53.85       108.15  \n",
       "3               No  Bank transfer (automatic)           42.30      1840.75  \n",
       "4              Yes           Electronic check           70.70       151.65  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "contracts_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "6c706d67",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>column_name</th>\n",
       "      <th>missing</th>\n",
       "      <th>%_missing</th>\n",
       "      <th>value_sample</th>\n",
       "      <th>dtype</th>\n",
       "      <th>count</th>\n",
       "      <th>unique</th>\n",
       "      <th>top</th>\n",
       "      <th>freq</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>min</th>\n",
       "      <th>25%</th>\n",
       "      <th>50%</th>\n",
       "      <th>75%</th>\n",
       "      <th>max</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>customerID</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>6356-ELRKD</td>\n",
       "      <td>object</td>\n",
       "      <td>7043</td>\n",
       "      <td>7043</td>\n",
       "      <td>5996-DAOQL</td>\n",
       "      <td>1</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>gender</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>Female</td>\n",
       "      <td>object</td>\n",
       "      <td>7043</td>\n",
       "      <td>2</td>\n",
       "      <td>Male</td>\n",
       "      <td>3555</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>SeniorCitizen</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>0</td>\n",
       "      <td>int64</td>\n",
       "      <td>7043</td>\n",
       "      <td>2</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td>0.162</td>\n",
       "      <td>0.369</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Partner</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>No</td>\n",
       "      <td>object</td>\n",
       "      <td>7043</td>\n",
       "      <td>2</td>\n",
       "      <td>No</td>\n",
       "      <td>3641</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Dependents</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>No</td>\n",
       "      <td>object</td>\n",
       "      <td>7043</td>\n",
       "      <td>2</td>\n",
       "      <td>No</td>\n",
       "      <td>4933</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     column_name  missing %_missing value_sample   dtype  count  unique  \\\n",
       "0     customerID        0    0.000%   6356-ELRKD  object   7043    7043   \n",
       "1         gender        0    0.000%       Female  object   7043       2   \n",
       "2  SeniorCitizen        0    0.000%            0   int64   7043       2   \n",
       "3        Partner        0    0.000%           No  object   7043       2   \n",
       "4     Dependents        0    0.000%           No  object   7043       2   \n",
       "\n",
       "          top  freq   mean    std  min  25%  50%  75%  max  \n",
       "0  5996-DAOQL     1                                         \n",
       "1        Male  3555                                         \n",
       "2                    0.162  0.369  0.0  0.0  0.0  0.0  1.0  \n",
       "3          No  3641                                         \n",
       "4          No  4933                                         "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "draw_general_info(personal_df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "061374e9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>customerID</th>\n",
       "      <th>gender</th>\n",
       "      <th>SeniorCitizen</th>\n",
       "      <th>Partner</th>\n",
       "      <th>Dependents</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>7590-VHVEG</td>\n",
       "      <td>Female</td>\n",
       "      <td>0</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>5575-GNVDE</td>\n",
       "      <td>Male</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3668-QPYBK</td>\n",
       "      <td>Male</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>7795-CFOCW</td>\n",
       "      <td>Male</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>9237-HQITU</td>\n",
       "      <td>Female</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   customerID  gender  SeniorCitizen Partner Dependents\n",
       "0  7590-VHVEG  Female              0     Yes         No\n",
       "1  5575-GNVDE    Male              0      No         No\n",
       "2  3668-QPYBK    Male              0      No         No\n",
       "3  7795-CFOCW    Male              0      No         No\n",
       "4  9237-HQITU  Female              0      No         No"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "personal_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "7d194593",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>column_name</th>\n",
       "      <th>missing</th>\n",
       "      <th>%_missing</th>\n",
       "      <th>value_sample</th>\n",
       "      <th>dtype</th>\n",
       "      <th>count</th>\n",
       "      <th>unique</th>\n",
       "      <th>top</th>\n",
       "      <th>freq</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>customerID</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>5855-EIBDE</td>\n",
       "      <td>object</td>\n",
       "      <td>5517</td>\n",
       "      <td>5517</td>\n",
       "      <td>1767-TGTKO</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>InternetService</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>Fiber optic</td>\n",
       "      <td>object</td>\n",
       "      <td>5517</td>\n",
       "      <td>2</td>\n",
       "      <td>Fiber optic</td>\n",
       "      <td>3096</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>OnlineSecurity</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>Yes</td>\n",
       "      <td>object</td>\n",
       "      <td>5517</td>\n",
       "      <td>2</td>\n",
       "      <td>No</td>\n",
       "      <td>3498</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>OnlineBackup</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>Yes</td>\n",
       "      <td>object</td>\n",
       "      <td>5517</td>\n",
       "      <td>2</td>\n",
       "      <td>No</td>\n",
       "      <td>3088</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>DeviceProtection</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>Yes</td>\n",
       "      <td>object</td>\n",
       "      <td>5517</td>\n",
       "      <td>2</td>\n",
       "      <td>No</td>\n",
       "      <td>3095</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>TechSupport</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>Yes</td>\n",
       "      <td>object</td>\n",
       "      <td>5517</td>\n",
       "      <td>2</td>\n",
       "      <td>No</td>\n",
       "      <td>3473</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>StreamingTV</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>No</td>\n",
       "      <td>object</td>\n",
       "      <td>5517</td>\n",
       "      <td>2</td>\n",
       "      <td>No</td>\n",
       "      <td>2810</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>StreamingMovies</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>Yes</td>\n",
       "      <td>object</td>\n",
       "      <td>5517</td>\n",
       "      <td>2</td>\n",
       "      <td>No</td>\n",
       "      <td>2785</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        column_name  missing %_missing value_sample   dtype  count  unique  \\\n",
       "0        customerID        0    0.000%   5855-EIBDE  object   5517    5517   \n",
       "1   InternetService        0    0.000%  Fiber optic  object   5517       2   \n",
       "2    OnlineSecurity        0    0.000%          Yes  object   5517       2   \n",
       "3      OnlineBackup        0    0.000%          Yes  object   5517       2   \n",
       "4  DeviceProtection        0    0.000%          Yes  object   5517       2   \n",
       "5       TechSupport        0    0.000%          Yes  object   5517       2   \n",
       "6       StreamingTV        0    0.000%           No  object   5517       2   \n",
       "7   StreamingMovies        0    0.000%          Yes  object   5517       2   \n",
       "\n",
       "           top  freq  \n",
       "0   1767-TGTKO     1  \n",
       "1  Fiber optic  3096  \n",
       "2           No  3498  \n",
       "3           No  3088  \n",
       "4           No  3095  \n",
       "5           No  3473  \n",
       "6           No  2810  \n",
       "7           No  2785  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "draw_general_info(internet_df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "88c5aaf6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>customerID</th>\n",
       "      <th>InternetService</th>\n",
       "      <th>OnlineSecurity</th>\n",
       "      <th>OnlineBackup</th>\n",
       "      <th>DeviceProtection</th>\n",
       "      <th>TechSupport</th>\n",
       "      <th>StreamingTV</th>\n",
       "      <th>StreamingMovies</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>7590-VHVEG</td>\n",
       "      <td>DSL</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>5575-GNVDE</td>\n",
       "      <td>DSL</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3668-QPYBK</td>\n",
       "      <td>DSL</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>7795-CFOCW</td>\n",
       "      <td>DSL</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>9237-HQITU</td>\n",
       "      <td>Fiber optic</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   customerID InternetService OnlineSecurity OnlineBackup DeviceProtection  \\\n",
       "0  7590-VHVEG             DSL             No          Yes               No   \n",
       "1  5575-GNVDE             DSL            Yes           No              Yes   \n",
       "2  3668-QPYBK             DSL            Yes          Yes               No   \n",
       "3  7795-CFOCW             DSL            Yes           No              Yes   \n",
       "4  9237-HQITU     Fiber optic             No           No               No   \n",
       "\n",
       "  TechSupport StreamingTV StreamingMovies  \n",
       "0          No          No              No  \n",
       "1          No          No              No  \n",
       "2          No          No              No  \n",
       "3         Yes          No              No  \n",
       "4          No          No              No  "
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "internet_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "3f430f72",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>column_name</th>\n",
       "      <th>missing</th>\n",
       "      <th>%_missing</th>\n",
       "      <th>value_sample</th>\n",
       "      <th>dtype</th>\n",
       "      <th>count</th>\n",
       "      <th>unique</th>\n",
       "      <th>top</th>\n",
       "      <th>freq</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>customerID</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>7416-CKTEP</td>\n",
       "      <td>object</td>\n",
       "      <td>6361</td>\n",
       "      <td>6361</td>\n",
       "      <td>5996-DAOQL</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>MultipleLines</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000%</td>\n",
       "      <td>No</td>\n",
       "      <td>object</td>\n",
       "      <td>6361</td>\n",
       "      <td>2</td>\n",
       "      <td>No</td>\n",
       "      <td>3390</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     column_name  missing %_missing value_sample   dtype  count  unique  \\\n",
       "0     customerID        0    0.000%   7416-CKTEP  object   6361    6361   \n",
       "1  MultipleLines        0    0.000%           No  object   6361       2   \n",
       "\n",
       "          top  freq  \n",
       "0  5996-DAOQL     1  \n",
       "1          No  3390  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "draw_general_info(phone_df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "a4f3c80e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>customerID</th>\n",
       "      <th>MultipleLines</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>5575-GNVDE</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3668-QPYBK</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>9237-HQITU</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>9305-CDSKC</td>\n",
       "      <td>Yes</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1452-KIOVK</td>\n",
       "      <td>Yes</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   customerID MultipleLines\n",
       "0  5575-GNVDE            No\n",
       "1  3668-QPYBK            No\n",
       "2  9237-HQITU            No\n",
       "3  9305-CDSKC           Yes\n",
       "4  1452-KIOVK           Yes"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "phone_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c80155dd",
   "metadata": {},
   "source": [
    "## Data Overview\n",
    "\n",
    "Our client telecom has tasked me with creating a model which can predict customer churn. \n",
    "\n",
    "Here are my initial notes on the data:\n",
    " \n",
    " - There are no missing values in this database\n",
    " - column names need to be changed to snakecase\n",
    " - Yes and No, male and female, dsl or fiberoptic, columns will be changed to binary: 1 or 0\n",
    " - begin date and end date should be changed to datetime data type\n",
    " \n",
    "## Project Plan\n",
    "\n",
    "1. Clean up the data as described above\n",
    "2. Merge the 4 dataframes across the Customer ID column \n",
    "3. Explore data in depth using EDA\n",
    "4. Use one hot encoding to engineer features. \n",
    "    - create a 'churned' column from the end date column, 1 for yes, 0 for no\n",
    "    - use one hot encoding on paymentmethod, type, and paperless billing\n",
    "5. Perform class balancing if discovered to be necessary in EDA\n",
    "6. Scale the data as necessary if indicated by EDA\n",
    "7. Drop non vital features\n",
    "8. Split the data into the train test and validation sets\n",
    "    - The target will be the engineered churn column\n",
    "9. Train the models and test them on the validation sets\n",
    "10. Tune models using the hyperparameters\n",
    "11. Choose the best scoring model and perform final test with the test set \n",
    "12. Write up the final conclusions \n",
    "\n",
    "I will be  using AUC_ROC and accuracy scores to test the models aptitude. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6829235e",
   "metadata": {},
   "source": [
    "## Data preprocessing  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "073edf12",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>customerID</th>\n",
       "      <th>BeginDate</th>\n",
       "      <th>EndDate</th>\n",
       "      <th>Type</th>\n",
       "      <th>PaperlessBilling</th>\n",
       "      <th>PaymentMethod</th>\n",
       "      <th>MonthlyCharges</th>\n",
       "      <th>TotalCharges</th>\n",
       "      <th>InternetService</th>\n",
       "      <th>OnlineSecurity</th>\n",
       "      <th>OnlineBackup</th>\n",
       "      <th>DeviceProtection</th>\n",
       "      <th>TechSupport</th>\n",
       "      <th>StreamingTV</th>\n",
       "      <th>StreamingMovies</th>\n",
       "      <th>gender</th>\n",
       "      <th>SeniorCitizen</th>\n",
       "      <th>Partner</th>\n",
       "      <th>Dependents</th>\n",
       "      <th>MultipleLines</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2061</th>\n",
       "      <td>4958-GZWIY</td>\n",
       "      <td>2019-04-01</td>\n",
       "      <td>2019-11-01 00:00:00</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>No</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>48.70</td>\n",
       "      <td>340.25</td>\n",
       "      <td>DSL</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Male</td>\n",
       "      <td>0</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2052</th>\n",
       "      <td>2038-LLMLM</td>\n",
       "      <td>2016-02-01</td>\n",
       "      <td>No</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>103.85</td>\n",
       "      <td>4946.05</td>\n",
       "      <td>Fiber optic</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Female</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6081</th>\n",
       "      <td>8757-TFHHJ</td>\n",
       "      <td>2020-01-01</td>\n",
       "      <td>No</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Credit card (automatic)</td>\n",
       "      <td>49.80</td>\n",
       "      <td>49.8</td>\n",
       "      <td>DSL</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Male</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2927</th>\n",
       "      <td>5542-TBBWB</td>\n",
       "      <td>2020-01-01</td>\n",
       "      <td>No</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>69.90</td>\n",
       "      <td>69.9</td>\n",
       "      <td>Fiber optic</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Male</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3081</th>\n",
       "      <td>6376-GAHQE</td>\n",
       "      <td>2019-07-01</td>\n",
       "      <td>2019-11-01 00:00:00</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>No</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>77.85</td>\n",
       "      <td>299.2</td>\n",
       "      <td>Fiber optic</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>Male</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      customerID   BeginDate              EndDate            Type  \\\n",
       "2061  4958-GZWIY  2019-04-01  2019-11-01 00:00:00  Month-to-month   \n",
       "2052  2038-LLMLM  2016-02-01                   No  Month-to-month   \n",
       "6081  8757-TFHHJ  2020-01-01                   No  Month-to-month   \n",
       "2927  5542-TBBWB  2020-01-01                   No  Month-to-month   \n",
       "3081  6376-GAHQE  2019-07-01  2019-11-01 00:00:00  Month-to-month   \n",
       "\n",
       "     PaperlessBilling            PaymentMethod  MonthlyCharges TotalCharges  \\\n",
       "2061               No         Electronic check           48.70       340.25   \n",
       "2052              Yes         Electronic check          103.85      4946.05   \n",
       "6081              Yes  Credit card (automatic)           49.80         49.8   \n",
       "2927              Yes         Electronic check           69.90         69.9   \n",
       "3081               No         Electronic check           77.85        299.2   \n",
       "\n",
       "     InternetService OnlineSecurity OnlineBackup DeviceProtection TechSupport  \\\n",
       "2061             DSL             No           No              Yes          No   \n",
       "2052     Fiber optic             No           No              Yes         Yes   \n",
       "6081             DSL            Yes           No               No          No   \n",
       "2927     Fiber optic             No           No               No          No   \n",
       "3081     Fiber optic             No           No               No          No   \n",
       "\n",
       "     StreamingTV StreamingMovies  gender  SeniorCitizen Partner Dependents  \\\n",
       "2061          No              No    Male              0     Yes        Yes   \n",
       "2052         Yes             Yes  Female              0      No         No   \n",
       "6081          No              No    Male              0      No         No   \n",
       "2927          No              No    Male              0      No         No   \n",
       "3081         Yes              No    Male              0      No         No   \n",
       "\n",
       "     MultipleLines  \n",
       "2061            No  \n",
       "2052           Yes  \n",
       "6081            No  \n",
       "2927            No  \n",
       "3081            No  "
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# joining datasets \n",
    "merged_df = pd.merge(contracts_df, internet_df, on=\"customerID\", how='left')\n",
    "merged_df1 = pd.merge(merged_df, personal_df, on=\"customerID\", how='left')\n",
    "merged_df2 = pd.merge(merged_df1, phone_df, on=\"customerID\", how='left')\n",
    "merged_df2 = merged_df2.fillna('No')\n",
    "merged_df2.sample(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "253747c3",
   "metadata": {},
   "outputs": [],
   "source": [
    "df=merged_df2.copy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "75d73e9a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 7043 entries, 0 to 7042\n",
      "Data columns (total 20 columns):\n",
      " #   Column            Non-Null Count  Dtype  \n",
      "---  ------            --------------  -----  \n",
      " 0   customerID        7043 non-null   object \n",
      " 1   BeginDate         7043 non-null   object \n",
      " 2   EndDate           7043 non-null   object \n",
      " 3   Type              7043 non-null   object \n",
      " 4   PaperlessBilling  7043 non-null   object \n",
      " 5   PaymentMethod     7043 non-null   object \n",
      " 6   MonthlyCharges    7043 non-null   float64\n",
      " 7   TotalCharges      7043 non-null   object \n",
      " 8   InternetService   7043 non-null   object \n",
      " 9   OnlineSecurity    7043 non-null   object \n",
      " 10  OnlineBackup      7043 non-null   object \n",
      " 11  DeviceProtection  7043 non-null   object \n",
      " 12  TechSupport       7043 non-null   object \n",
      " 13  StreamingTV       7043 non-null   object \n",
      " 14  StreamingMovies   7043 non-null   object \n",
      " 15  gender            7043 non-null   object \n",
      " 16  SeniorCitizen     7043 non-null   int64  \n",
      " 17  Partner           7043 non-null   object \n",
      " 18  Dependents        7043 non-null   object \n",
      " 19  MultipleLines     7043 non-null   object \n",
      "dtypes: float64(1), int64(1), object(18)\n",
      "memory usage: 1.1+ MB\n"
     ]
    }
   ],
   "source": [
    "df.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "f8e42053",
   "metadata": {},
   "outputs": [],
   "source": [
    "#Changing the column names to snake_case\n",
    "\n",
    "df.columns = [x.lower() for x in df.columns]\n",
    "df.columns = df.columns.str.replace(\"[ ]\", \"_\", regex=True)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "9203128c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>customerid</th>\n",
       "      <th>begindate</th>\n",
       "      <th>enddate</th>\n",
       "      <th>type</th>\n",
       "      <th>paperlessbilling</th>\n",
       "      <th>paymentmethod</th>\n",
       "      <th>monthlycharges</th>\n",
       "      <th>totalcharges</th>\n",
       "      <th>internetservice</th>\n",
       "      <th>onlinesecurity</th>\n",
       "      <th>onlinebackup</th>\n",
       "      <th>deviceprotection</th>\n",
       "      <th>techsupport</th>\n",
       "      <th>streamingtv</th>\n",
       "      <th>streamingmovies</th>\n",
       "      <th>gender</th>\n",
       "      <th>seniorcitizen</th>\n",
       "      <th>partner</th>\n",
       "      <th>dependents</th>\n",
       "      <th>multiplelines</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>7590-VHVEG</td>\n",
       "      <td>2020-01-01</td>\n",
       "      <td>No</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>29.85</td>\n",
       "      <td>29.85</td>\n",
       "      <td>DSL</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Female</td>\n",
       "      <td>0</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>5575-GNVDE</td>\n",
       "      <td>2017-04-01</td>\n",
       "      <td>No</td>\n",
       "      <td>One year</td>\n",
       "      <td>No</td>\n",
       "      <td>Mailed check</td>\n",
       "      <td>56.95</td>\n",
       "      <td>1889.5</td>\n",
       "      <td>DSL</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Male</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3668-QPYBK</td>\n",
       "      <td>2019-10-01</td>\n",
       "      <td>2019-12-01 00:00:00</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Mailed check</td>\n",
       "      <td>53.85</td>\n",
       "      <td>108.15</td>\n",
       "      <td>DSL</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Male</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>7795-CFOCW</td>\n",
       "      <td>2016-05-01</td>\n",
       "      <td>No</td>\n",
       "      <td>One year</td>\n",
       "      <td>No</td>\n",
       "      <td>Bank transfer (automatic)</td>\n",
       "      <td>42.30</td>\n",
       "      <td>1840.75</td>\n",
       "      <td>DSL</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Male</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>9237-HQITU</td>\n",
       "      <td>2019-09-01</td>\n",
       "      <td>2019-11-01 00:00:00</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>70.70</td>\n",
       "      <td>151.65</td>\n",
       "      <td>Fiber optic</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Female</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   customerid   begindate              enddate            type  \\\n",
       "0  7590-VHVEG  2020-01-01                   No  Month-to-month   \n",
       "1  5575-GNVDE  2017-04-01                   No        One year   \n",
       "2  3668-QPYBK  2019-10-01  2019-12-01 00:00:00  Month-to-month   \n",
       "3  7795-CFOCW  2016-05-01                   No        One year   \n",
       "4  9237-HQITU  2019-09-01  2019-11-01 00:00:00  Month-to-month   \n",
       "\n",
       "  paperlessbilling              paymentmethod  monthlycharges totalcharges  \\\n",
       "0              Yes           Electronic check           29.85        29.85   \n",
       "1               No               Mailed check           56.95       1889.5   \n",
       "2              Yes               Mailed check           53.85       108.15   \n",
       "3               No  Bank transfer (automatic)           42.30      1840.75   \n",
       "4              Yes           Electronic check           70.70       151.65   \n",
       "\n",
       "  internetservice onlinesecurity onlinebackup deviceprotection techsupport  \\\n",
       "0             DSL             No          Yes               No          No   \n",
       "1             DSL            Yes           No              Yes          No   \n",
       "2             DSL            Yes          Yes               No          No   \n",
       "3             DSL            Yes           No              Yes         Yes   \n",
       "4     Fiber optic             No           No               No          No   \n",
       "\n",
       "  streamingtv streamingmovies  gender  seniorcitizen partner dependents  \\\n",
       "0          No              No  Female              0     Yes         No   \n",
       "1          No              No    Male              0      No         No   \n",
       "2          No              No    Male              0      No         No   \n",
       "3          No              No    Male              0      No         No   \n",
       "4          No              No  Female              0      No         No   \n",
       "\n",
       "  multiplelines  \n",
       "0            No  \n",
       "1            No  \n",
       "2            No  \n",
       "3            No  \n",
       "4            No  "
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "84659cf4",
   "metadata": {},
   "outputs": [],
   "source": [
    "df= df.rename(columns={'customerid' : 'customer_id',\n",
    "                   'begindate' : 'begin_date',\n",
    "                   'enddate' : 'end_date',\n",
    "                   'paperlessbilling' : 'paperless_billing',\n",
    "                   'paymentmethod' : 'payment_method',\n",
    "                   'monthlycharges' : 'monthly_charges',\n",
    "                   'totalcharges' : 'total_charges',\n",
    "                   'seniorcitizen' : 'senior_citizen',\n",
    "                   'internetservice' : 'internet_service',\n",
    "                   'onlinesecurity' : 'online_security',\n",
    "                   'onlinebackup' : 'online_backup',\n",
    "                   'deviceprotection' : 'device_protection',\n",
    "                   'techsupport' : 'tech_support',\n",
    "                   'streamingtv' : 'streaming_tv',\n",
    "                   'streamingmovies' : 'streaming_movies',\n",
    "                   'multiplelines' : 'multiple_lines'}\n",
    "         )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "dca8fd0d",
   "metadata": {},
   "outputs": [],
   "source": [
    "df.gender.replace(('Male', 'Female'), (1, 0), inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "f7f4f560",
   "metadata": {},
   "outputs": [],
   "source": [
    "df.internet_service.replace(('DSL', 'Fiber optic'), (1, 0), inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "273a2709",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>customer_id</th>\n",
       "      <th>begin_date</th>\n",
       "      <th>end_date</th>\n",
       "      <th>type</th>\n",
       "      <th>paperless_billing</th>\n",
       "      <th>payment_method</th>\n",
       "      <th>monthly_charges</th>\n",
       "      <th>total_charges</th>\n",
       "      <th>internet_service</th>\n",
       "      <th>online_security</th>\n",
       "      <th>online_backup</th>\n",
       "      <th>device_protection</th>\n",
       "      <th>tech_support</th>\n",
       "      <th>streaming_tv</th>\n",
       "      <th>streaming_movies</th>\n",
       "      <th>gender</th>\n",
       "      <th>senior_citizen</th>\n",
       "      <th>partner</th>\n",
       "      <th>dependents</th>\n",
       "      <th>multiple_lines</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>7590-VHVEG</td>\n",
       "      <td>2020-01-01</td>\n",
       "      <td>No</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>29.85</td>\n",
       "      <td>29.85</td>\n",
       "      <td>1</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>5575-GNVDE</td>\n",
       "      <td>2017-04-01</td>\n",
       "      <td>No</td>\n",
       "      <td>One year</td>\n",
       "      <td>No</td>\n",
       "      <td>Mailed check</td>\n",
       "      <td>56.95</td>\n",
       "      <td>1889.5</td>\n",
       "      <td>1</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3668-QPYBK</td>\n",
       "      <td>2019-10-01</td>\n",
       "      <td>2019-12-01 00:00:00</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Mailed check</td>\n",
       "      <td>53.85</td>\n",
       "      <td>108.15</td>\n",
       "      <td>1</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>7795-CFOCW</td>\n",
       "      <td>2016-05-01</td>\n",
       "      <td>No</td>\n",
       "      <td>One year</td>\n",
       "      <td>No</td>\n",
       "      <td>Bank transfer (automatic)</td>\n",
       "      <td>42.30</td>\n",
       "      <td>1840.75</td>\n",
       "      <td>1</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>9237-HQITU</td>\n",
       "      <td>2019-09-01</td>\n",
       "      <td>2019-11-01 00:00:00</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>70.70</td>\n",
       "      <td>151.65</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  customer_id  begin_date             end_date            type  \\\n",
       "0  7590-VHVEG  2020-01-01                   No  Month-to-month   \n",
       "1  5575-GNVDE  2017-04-01                   No        One year   \n",
       "2  3668-QPYBK  2019-10-01  2019-12-01 00:00:00  Month-to-month   \n",
       "3  7795-CFOCW  2016-05-01                   No        One year   \n",
       "4  9237-HQITU  2019-09-01  2019-11-01 00:00:00  Month-to-month   \n",
       "\n",
       "  paperless_billing             payment_method  monthly_charges total_charges  \\\n",
       "0               Yes           Electronic check            29.85         29.85   \n",
       "1                No               Mailed check            56.95        1889.5   \n",
       "2               Yes               Mailed check            53.85        108.15   \n",
       "3                No  Bank transfer (automatic)            42.30       1840.75   \n",
       "4               Yes           Electronic check            70.70        151.65   \n",
       "\n",
       "  internet_service online_security online_backup device_protection  \\\n",
       "0                1              No           Yes                No   \n",
       "1                1             Yes            No               Yes   \n",
       "2                1             Yes           Yes                No   \n",
       "3                1             Yes            No               Yes   \n",
       "4                0              No            No                No   \n",
       "\n",
       "  tech_support streaming_tv streaming_movies  gender  senior_citizen partner  \\\n",
       "0           No           No               No       0               0     Yes   \n",
       "1           No           No               No       1               0      No   \n",
       "2           No           No               No       1               0      No   \n",
       "3          Yes           No               No       1               0      No   \n",
       "4           No           No               No       0               0      No   \n",
       "\n",
       "  dependents multiple_lines  \n",
       "0         No             No  \n",
       "1         No             No  \n",
       "2         No             No  \n",
       "3         No             No  \n",
       "4         No             No  "
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "005be1b1",
   "metadata": {},
   "outputs": [],
   "source": [
    "#df.replace(('Yes', 'No'), (1, 0), inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "c6a61938",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>customer_id</th>\n",
       "      <th>begin_date</th>\n",
       "      <th>end_date</th>\n",
       "      <th>type</th>\n",
       "      <th>paperless_billing</th>\n",
       "      <th>payment_method</th>\n",
       "      <th>monthly_charges</th>\n",
       "      <th>total_charges</th>\n",
       "      <th>internet_service</th>\n",
       "      <th>online_security</th>\n",
       "      <th>online_backup</th>\n",
       "      <th>device_protection</th>\n",
       "      <th>tech_support</th>\n",
       "      <th>streaming_tv</th>\n",
       "      <th>streaming_movies</th>\n",
       "      <th>gender</th>\n",
       "      <th>senior_citizen</th>\n",
       "      <th>partner</th>\n",
       "      <th>dependents</th>\n",
       "      <th>multiple_lines</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>7590-VHVEG</td>\n",
       "      <td>2020-01-01</td>\n",
       "      <td>No</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>29.85</td>\n",
       "      <td>29.85</td>\n",
       "      <td>1</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>5575-GNVDE</td>\n",
       "      <td>2017-04-01</td>\n",
       "      <td>No</td>\n",
       "      <td>One year</td>\n",
       "      <td>No</td>\n",
       "      <td>Mailed check</td>\n",
       "      <td>56.95</td>\n",
       "      <td>1889.5</td>\n",
       "      <td>1</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3668-QPYBK</td>\n",
       "      <td>2019-10-01</td>\n",
       "      <td>2019-12-01 00:00:00</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Mailed check</td>\n",
       "      <td>53.85</td>\n",
       "      <td>108.15</td>\n",
       "      <td>1</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>7795-CFOCW</td>\n",
       "      <td>2016-05-01</td>\n",
       "      <td>No</td>\n",
       "      <td>One year</td>\n",
       "      <td>No</td>\n",
       "      <td>Bank transfer (automatic)</td>\n",
       "      <td>42.30</td>\n",
       "      <td>1840.75</td>\n",
       "      <td>1</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>9237-HQITU</td>\n",
       "      <td>2019-09-01</td>\n",
       "      <td>2019-11-01 00:00:00</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>70.70</td>\n",
       "      <td>151.65</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  customer_id  begin_date             end_date            type  \\\n",
       "0  7590-VHVEG  2020-01-01                   No  Month-to-month   \n",
       "1  5575-GNVDE  2017-04-01                   No        One year   \n",
       "2  3668-QPYBK  2019-10-01  2019-12-01 00:00:00  Month-to-month   \n",
       "3  7795-CFOCW  2016-05-01                   No        One year   \n",
       "4  9237-HQITU  2019-09-01  2019-11-01 00:00:00  Month-to-month   \n",
       "\n",
       "  paperless_billing             payment_method  monthly_charges total_charges  \\\n",
       "0               Yes           Electronic check            29.85         29.85   \n",
       "1                No               Mailed check            56.95        1889.5   \n",
       "2               Yes               Mailed check            53.85        108.15   \n",
       "3                No  Bank transfer (automatic)            42.30       1840.75   \n",
       "4               Yes           Electronic check            70.70        151.65   \n",
       "\n",
       "  internet_service online_security online_backup device_protection  \\\n",
       "0                1              No           Yes                No   \n",
       "1                1             Yes            No               Yes   \n",
       "2                1             Yes           Yes                No   \n",
       "3                1             Yes            No               Yes   \n",
       "4                0              No            No                No   \n",
       "\n",
       "  tech_support streaming_tv streaming_movies  gender  senior_citizen partner  \\\n",
       "0           No           No               No       0               0     Yes   \n",
       "1           No           No               No       1               0      No   \n",
       "2           No           No               No       1               0      No   \n",
       "3          Yes           No               No       1               0      No   \n",
       "4           No           No               No       0               0      No   \n",
       "\n",
       "  dependents multiple_lines  \n",
       "0         No             No  \n",
       "1         No             No  \n",
       "2         No             No  \n",
       "3         No             No  \n",
       "4         No             No  "
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "9d00180f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>monthly_charges</th>\n",
       "      <th>gender</th>\n",
       "      <th>senior_citizen</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>7043.000000</td>\n",
       "      <td>7043.000000</td>\n",
       "      <td>7043.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>64.761692</td>\n",
       "      <td>0.504756</td>\n",
       "      <td>0.162147</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>30.090047</td>\n",
       "      <td>0.500013</td>\n",
       "      <td>0.368612</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>18.250000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>35.500000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>70.350000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>89.850000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>118.750000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       monthly_charges       gender  senior_citizen\n",
       "count      7043.000000  7043.000000     7043.000000\n",
       "mean         64.761692     0.504756        0.162147\n",
       "std          30.090047     0.500013        0.368612\n",
       "min          18.250000     0.000000        0.000000\n",
       "25%          35.500000     0.000000        0.000000\n",
       "50%          70.350000     1.000000        0.000000\n",
       "75%          89.850000     1.000000        0.000000\n",
       "max         118.750000     1.000000        1.000000"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "aa69a85c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 7043 entries, 0 to 7042\n",
      "Data columns (total 20 columns):\n",
      " #   Column             Non-Null Count  Dtype  \n",
      "---  ------             --------------  -----  \n",
      " 0   customer_id        7043 non-null   object \n",
      " 1   begin_date         7043 non-null   object \n",
      " 2   end_date           7043 non-null   object \n",
      " 3   type               7043 non-null   object \n",
      " 4   paperless_billing  7043 non-null   object \n",
      " 5   payment_method     7043 non-null   object \n",
      " 6   monthly_charges    7043 non-null   float64\n",
      " 7   total_charges      7043 non-null   object \n",
      " 8   internet_service   7043 non-null   object \n",
      " 9   online_security    7043 non-null   object \n",
      " 10  online_backup      7043 non-null   object \n",
      " 11  device_protection  7043 non-null   object \n",
      " 12  tech_support       7043 non-null   object \n",
      " 13  streaming_tv       7043 non-null   object \n",
      " 14  streaming_movies   7043 non-null   object \n",
      " 15  gender             7043 non-null   int64  \n",
      " 16  senior_citizen     7043 non-null   int64  \n",
      " 17  partner            7043 non-null   object \n",
      " 18  dependents         7043 non-null   object \n",
      " 19  multiple_lines     7043 non-null   object \n",
      "dtypes: float64(1), int64(2), object(17)\n",
      "memory usage: 1.1+ MB\n"
     ]
    }
   ],
   "source": [
    "df.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "f50d2db0",
   "metadata": {},
   "outputs": [],
   "source": [
    "def change_datatype(df, cols, type_val):\n",
    "    for col in cols:\n",
    "        df[col] = df[col].astype(type_val)\n",
    "\n",
    "# create new end date feature\n",
    "list_value = []\n",
    "for value in df.end_date:\n",
    "    if value != 'No':\n",
    "        datetime_value = datetime.strptime(value, '%Y-%m-%d %H:%M:%S')\n",
    "        list_value.append(datetime_value)\n",
    "    else:\n",
    "        contract_valid_date = '2020-02-01 00:00:00' # date contract is valid\n",
    "        contract_valid_date_value = datetime.strptime(contract_valid_date, '%Y-%m-%d %H:%M:%S')\n",
    "        list_value.append(contract_valid_date_value)\n",
    "EndDate_value = pd.to_datetime(list_value)\n",
    "df.insert(3, 'end_date_value', EndDate_value)\n",
    "\n",
    "# prepare TotalChargers \n",
    "df.loc[df['total_charges'].isin([' ']),'total_charges'] = 0\n",
    "\n",
    "# change datatypes\n",
    "change_datatype(df, ['begin_date'], 'datetime64[ns]')\n",
    "change_datatype(df, ['monthly_charges', 'total_charges'], 'float32')\n",
    "change_datatype(df, ['senior_citizen'], 'int32')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "cd895ea7",
   "metadata": {},
   "outputs": [],
   "source": [
    "#changing the data types\n",
    "df['begin_date'] = pd.to_datetime(df['begin_date'],format='%Y%m%d', errors='ignore')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "ca59d015",
   "metadata": {},
   "outputs": [],
   "source": [
    "df['end_date'] = pd.to_datetime(df['end_date'],format='%Y%m%d', errors='ignore')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "5492b83c",
   "metadata": {},
   "outputs": [],
   "source": [
    "df.loc[df['total_charges'].isin([' ']),'total_charges'] = 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "e4e3717c",
   "metadata": {},
   "outputs": [],
   "source": [
    "df['total_charges'] = df['total_charges'].astype('float')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "aa43d67e",
   "metadata": {},
   "outputs": [],
   "source": [
    "df['senior_citizen'] = df['senior_citizen'].astype('int64')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "60307555",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 7043 entries, 0 to 7042\n",
      "Data columns (total 21 columns):\n",
      " #   Column             Non-Null Count  Dtype         \n",
      "---  ------             --------------  -----         \n",
      " 0   customer_id        7043 non-null   object        \n",
      " 1   begin_date         7043 non-null   datetime64[ns]\n",
      " 2   end_date           7043 non-null   object        \n",
      " 3   end_date_value     7043 non-null   datetime64[ns]\n",
      " 4   type               7043 non-null   object        \n",
      " 5   paperless_billing  7043 non-null   object        \n",
      " 6   payment_method     7043 non-null   object        \n",
      " 7   monthly_charges    7043 non-null   float32       \n",
      " 8   total_charges      7043 non-null   float64       \n",
      " 9   internet_service   7043 non-null   object        \n",
      " 10  online_security    7043 non-null   object        \n",
      " 11  online_backup      7043 non-null   object        \n",
      " 12  device_protection  7043 non-null   object        \n",
      " 13  tech_support       7043 non-null   object        \n",
      " 14  streaming_tv       7043 non-null   object        \n",
      " 15  streaming_movies   7043 non-null   object        \n",
      " 16  gender             7043 non-null   int64         \n",
      " 17  senior_citizen     7043 non-null   int64         \n",
      " 18  partner            7043 non-null   object        \n",
      " 19  dependents         7043 non-null   object        \n",
      " 20  multiple_lines     7043 non-null   object        \n",
      "dtypes: datetime64[ns](2), float32(1), float64(1), int64(2), object(15)\n",
      "memory usage: 1.2+ MB\n"
     ]
    }
   ],
   "source": [
    "df.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "a3db327d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>customer_id</th>\n",
       "      <th>begin_date</th>\n",
       "      <th>end_date</th>\n",
       "      <th>end_date_value</th>\n",
       "      <th>type</th>\n",
       "      <th>paperless_billing</th>\n",
       "      <th>payment_method</th>\n",
       "      <th>monthly_charges</th>\n",
       "      <th>total_charges</th>\n",
       "      <th>internet_service</th>\n",
       "      <th>...</th>\n",
       "      <th>online_backup</th>\n",
       "      <th>device_protection</th>\n",
       "      <th>tech_support</th>\n",
       "      <th>streaming_tv</th>\n",
       "      <th>streaming_movies</th>\n",
       "      <th>gender</th>\n",
       "      <th>senior_citizen</th>\n",
       "      <th>partner</th>\n",
       "      <th>dependents</th>\n",
       "      <th>multiple_lines</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>7590-VHVEG</td>\n",
       "      <td>2020-01-01</td>\n",
       "      <td>No</td>\n",
       "      <td>2020-02-01</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>29.850000</td>\n",
       "      <td>29.850000</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>5575-GNVDE</td>\n",
       "      <td>2017-04-01</td>\n",
       "      <td>No</td>\n",
       "      <td>2020-02-01</td>\n",
       "      <td>One year</td>\n",
       "      <td>No</td>\n",
       "      <td>Mailed check</td>\n",
       "      <td>56.950001</td>\n",
       "      <td>1889.500000</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3668-QPYBK</td>\n",
       "      <td>2019-10-01</td>\n",
       "      <td>2019-12-01 00:00:00</td>\n",
       "      <td>2019-12-01</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Mailed check</td>\n",
       "      <td>53.849998</td>\n",
       "      <td>108.150002</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>7795-CFOCW</td>\n",
       "      <td>2016-05-01</td>\n",
       "      <td>No</td>\n",
       "      <td>2020-02-01</td>\n",
       "      <td>One year</td>\n",
       "      <td>No</td>\n",
       "      <td>Bank transfer (automatic)</td>\n",
       "      <td>42.299999</td>\n",
       "      <td>1840.750000</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>9237-HQITU</td>\n",
       "      <td>2019-09-01</td>\n",
       "      <td>2019-11-01 00:00:00</td>\n",
       "      <td>2019-11-01</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>70.699997</td>\n",
       "      <td>151.649994</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 21 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "  customer_id begin_date             end_date end_date_value            type  \\\n",
       "0  7590-VHVEG 2020-01-01                   No     2020-02-01  Month-to-month   \n",
       "1  5575-GNVDE 2017-04-01                   No     2020-02-01        One year   \n",
       "2  3668-QPYBK 2019-10-01  2019-12-01 00:00:00     2019-12-01  Month-to-month   \n",
       "3  7795-CFOCW 2016-05-01                   No     2020-02-01        One year   \n",
       "4  9237-HQITU 2019-09-01  2019-11-01 00:00:00     2019-11-01  Month-to-month   \n",
       "\n",
       "  paperless_billing             payment_method  monthly_charges  \\\n",
       "0               Yes           Electronic check        29.850000   \n",
       "1                No               Mailed check        56.950001   \n",
       "2               Yes               Mailed check        53.849998   \n",
       "3                No  Bank transfer (automatic)        42.299999   \n",
       "4               Yes           Electronic check        70.699997   \n",
       "\n",
       "   total_charges internet_service  ... online_backup device_protection  \\\n",
       "0      29.850000                1  ...           Yes                No   \n",
       "1    1889.500000                1  ...            No               Yes   \n",
       "2     108.150002                1  ...           Yes                No   \n",
       "3    1840.750000                1  ...            No               Yes   \n",
       "4     151.649994                0  ...            No                No   \n",
       "\n",
       "  tech_support streaming_tv streaming_movies gender  senior_citizen  partner  \\\n",
       "0           No           No               No      0               0      Yes   \n",
       "1           No           No               No      1               0       No   \n",
       "2           No           No               No      1               0       No   \n",
       "3          Yes           No               No      1               0       No   \n",
       "4           No           No               No      0               0       No   \n",
       "\n",
       "  dependents multiple_lines  \n",
       "0         No             No  \n",
       "1         No             No  \n",
       "2         No             No  \n",
       "3         No             No  \n",
       "4         No             No  \n",
       "\n",
       "[5 rows x 21 columns]"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "730342d4",
   "metadata": {},
   "source": [
    "## Feature Engineering \n",
    "I will engineer features for length of service as 'tenure', a binary column for if a customer has churned or not as 'churned', a column for the number of services a person is using, a binary column for if a person has a credit card on file as 'has_card', and columns for the 'year', 'month', and 'dayofweek' the customer joined telecom "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "40a18dfd",
   "metadata": {},
   "outputs": [],
   "source": [
    "# change date type to datetime and split into day, month and year\n",
    "def new_date_features(df):\n",
    "    columns = df.columns.tolist()\n",
    "    idx = [columns.index(x) for x in columns if 'begin_date' in x][0]\n",
    "    \n",
    "    df[columns[idx]] = pd.to_datetime(df[columns[idx]])\n",
    "    df['dayofweek'] = df[columns[idx]].dt.day_name()\n",
    "    df['month'] = df[columns[idx]].dt.month_name()\n",
    "    df['year'] = df[columns[idx]].dt.year\n",
    "    return df; "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "cd303f34",
   "metadata": {},
   "outputs": [],
   "source": [
    "new_date_features(df)\n",
    "df['tenure'] = df['end_date_value'].dt.year - df['begin_date'].dt.year\n",
    "df['has_crcard'] = [1 if x == 'Credit card (automatic)' else 0 for x in df['payment_method']]\n",
    "df['exited'] = [1 if x != 'No' else 0 for x in df['end_date']]\n",
    "df['service_count'] = [x.count('Yes') for x in zip(df['online_security'], df['online_backup'], df['device_protection'], \n",
    "                                                           df['tech_support'], df['streaming_tv'], df['streaming_movies'])]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "6870b6ba",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>customer_id</th>\n",
       "      <th>begin_date</th>\n",
       "      <th>end_date</th>\n",
       "      <th>end_date_value</th>\n",
       "      <th>type</th>\n",
       "      <th>paperless_billing</th>\n",
       "      <th>payment_method</th>\n",
       "      <th>monthly_charges</th>\n",
       "      <th>total_charges</th>\n",
       "      <th>internet_service</th>\n",
       "      <th>...</th>\n",
       "      <th>partner</th>\n",
       "      <th>dependents</th>\n",
       "      <th>multiple_lines</th>\n",
       "      <th>dayofweek</th>\n",
       "      <th>month</th>\n",
       "      <th>year</th>\n",
       "      <th>tenure</th>\n",
       "      <th>has_crcard</th>\n",
       "      <th>exited</th>\n",
       "      <th>service_count</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>7590-VHVEG</td>\n",
       "      <td>2020-01-01</td>\n",
       "      <td>No</td>\n",
       "      <td>2020-02-01</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>29.850000</td>\n",
       "      <td>29.850000</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Wednesday</td>\n",
       "      <td>January</td>\n",
       "      <td>2020</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>5575-GNVDE</td>\n",
       "      <td>2017-04-01</td>\n",
       "      <td>No</td>\n",
       "      <td>2020-02-01</td>\n",
       "      <td>One year</td>\n",
       "      <td>No</td>\n",
       "      <td>Mailed check</td>\n",
       "      <td>56.950001</td>\n",
       "      <td>1889.500000</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Saturday</td>\n",
       "      <td>April</td>\n",
       "      <td>2017</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3668-QPYBK</td>\n",
       "      <td>2019-10-01</td>\n",
       "      <td>2019-12-01 00:00:00</td>\n",
       "      <td>2019-12-01</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Mailed check</td>\n",
       "      <td>53.849998</td>\n",
       "      <td>108.150002</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Tuesday</td>\n",
       "      <td>October</td>\n",
       "      <td>2019</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>7795-CFOCW</td>\n",
       "      <td>2016-05-01</td>\n",
       "      <td>No</td>\n",
       "      <td>2020-02-01</td>\n",
       "      <td>One year</td>\n",
       "      <td>No</td>\n",
       "      <td>Bank transfer (automatic)</td>\n",
       "      <td>42.299999</td>\n",
       "      <td>1840.750000</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Sunday</td>\n",
       "      <td>May</td>\n",
       "      <td>2016</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>9237-HQITU</td>\n",
       "      <td>2019-09-01</td>\n",
       "      <td>2019-11-01 00:00:00</td>\n",
       "      <td>2019-11-01</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>70.699997</td>\n",
       "      <td>151.649994</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Sunday</td>\n",
       "      <td>September</td>\n",
       "      <td>2019</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>9305-CDSKC</td>\n",
       "      <td>2019-03-01</td>\n",
       "      <td>2019-11-01 00:00:00</td>\n",
       "      <td>2019-11-01</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>99.650002</td>\n",
       "      <td>820.500000</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Friday</td>\n",
       "      <td>March</td>\n",
       "      <td>2019</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>1452-KIOVK</td>\n",
       "      <td>2018-04-01</td>\n",
       "      <td>No</td>\n",
       "      <td>2020-02-01</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Credit card (automatic)</td>\n",
       "      <td>89.099998</td>\n",
       "      <td>1949.400024</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Sunday</td>\n",
       "      <td>April</td>\n",
       "      <td>2018</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>6713-OKOMC</td>\n",
       "      <td>2019-04-01</td>\n",
       "      <td>No</td>\n",
       "      <td>2020-02-01</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>No</td>\n",
       "      <td>Mailed check</td>\n",
       "      <td>29.750000</td>\n",
       "      <td>301.899994</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Monday</td>\n",
       "      <td>April</td>\n",
       "      <td>2019</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>7892-POOKP</td>\n",
       "      <td>2017-07-01</td>\n",
       "      <td>2019-11-01 00:00:00</td>\n",
       "      <td>2019-11-01</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>104.800003</td>\n",
       "      <td>3046.050049</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Saturday</td>\n",
       "      <td>July</td>\n",
       "      <td>2017</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>6388-TABGU</td>\n",
       "      <td>2014-12-01</td>\n",
       "      <td>No</td>\n",
       "      <td>2020-02-01</td>\n",
       "      <td>One year</td>\n",
       "      <td>No</td>\n",
       "      <td>Bank transfer (automatic)</td>\n",
       "      <td>56.150002</td>\n",
       "      <td>3487.949951</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>Monday</td>\n",
       "      <td>December</td>\n",
       "      <td>2014</td>\n",
       "      <td>6</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>9763-GRSKD</td>\n",
       "      <td>2019-01-01</td>\n",
       "      <td>No</td>\n",
       "      <td>2020-02-01</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Mailed check</td>\n",
       "      <td>49.950001</td>\n",
       "      <td>587.450012</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>Tuesday</td>\n",
       "      <td>January</td>\n",
       "      <td>2019</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>7469-LKBCI</td>\n",
       "      <td>2018-10-01</td>\n",
       "      <td>No</td>\n",
       "      <td>2020-02-01</td>\n",
       "      <td>Two year</td>\n",
       "      <td>No</td>\n",
       "      <td>Credit card (automatic)</td>\n",
       "      <td>18.950001</td>\n",
       "      <td>326.799988</td>\n",
       "      <td>No</td>\n",
       "      <td>...</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Monday</td>\n",
       "      <td>October</td>\n",
       "      <td>2018</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>8091-TTVAX</td>\n",
       "      <td>2015-04-01</td>\n",
       "      <td>No</td>\n",
       "      <td>2020-02-01</td>\n",
       "      <td>One year</td>\n",
       "      <td>No</td>\n",
       "      <td>Credit card (automatic)</td>\n",
       "      <td>100.349998</td>\n",
       "      <td>5681.100098</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Wednesday</td>\n",
       "      <td>April</td>\n",
       "      <td>2015</td>\n",
       "      <td>5</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>0280-XJGEX</td>\n",
       "      <td>2015-09-01</td>\n",
       "      <td>2019-10-01 00:00:00</td>\n",
       "      <td>2019-10-01</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Bank transfer (automatic)</td>\n",
       "      <td>103.699997</td>\n",
       "      <td>5036.299805</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Tuesday</td>\n",
       "      <td>September</td>\n",
       "      <td>2015</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>5129-JLPIS</td>\n",
       "      <td>2018-01-01</td>\n",
       "      <td>No</td>\n",
       "      <td>2020-02-01</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>105.500000</td>\n",
       "      <td>2686.050049</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Monday</td>\n",
       "      <td>January</td>\n",
       "      <td>2018</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>3655-SNQYZ</td>\n",
       "      <td>2014-05-01</td>\n",
       "      <td>No</td>\n",
       "      <td>2020-02-01</td>\n",
       "      <td>Two year</td>\n",
       "      <td>No</td>\n",
       "      <td>Credit card (automatic)</td>\n",
       "      <td>113.250000</td>\n",
       "      <td>7895.149902</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Thursday</td>\n",
       "      <td>May</td>\n",
       "      <td>2014</td>\n",
       "      <td>6</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>8191-XWSZG</td>\n",
       "      <td>2015-10-01</td>\n",
       "      <td>No</td>\n",
       "      <td>2020-02-01</td>\n",
       "      <td>One year</td>\n",
       "      <td>No</td>\n",
       "      <td>Mailed check</td>\n",
       "      <td>20.650000</td>\n",
       "      <td>1022.950012</td>\n",
       "      <td>No</td>\n",
       "      <td>...</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Thursday</td>\n",
       "      <td>October</td>\n",
       "      <td>2015</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>9959-WOFKT</td>\n",
       "      <td>2014-03-01</td>\n",
       "      <td>No</td>\n",
       "      <td>2020-02-01</td>\n",
       "      <td>Two year</td>\n",
       "      <td>No</td>\n",
       "      <td>Bank transfer (automatic)</td>\n",
       "      <td>106.699997</td>\n",
       "      <td>7382.250000</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Saturday</td>\n",
       "      <td>March</td>\n",
       "      <td>2014</td>\n",
       "      <td>6</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>4190-MFLUW</td>\n",
       "      <td>2019-01-01</td>\n",
       "      <td>2019-11-01 00:00:00</td>\n",
       "      <td>2019-11-01</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>No</td>\n",
       "      <td>Credit card (automatic)</td>\n",
       "      <td>55.200001</td>\n",
       "      <td>528.349976</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>No</td>\n",
       "      <td>Tuesday</td>\n",
       "      <td>January</td>\n",
       "      <td>2019</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>4183-MYFRB</td>\n",
       "      <td>2018-05-01</td>\n",
       "      <td>No</td>\n",
       "      <td>2020-02-01</td>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Electronic check</td>\n",
       "      <td>90.050003</td>\n",
       "      <td>1862.900024</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Tuesday</td>\n",
       "      <td>May</td>\n",
       "      <td>2018</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>20 rows × 28 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   customer_id begin_date             end_date end_date_value            type  \\\n",
       "0   7590-VHVEG 2020-01-01                   No     2020-02-01  Month-to-month   \n",
       "1   5575-GNVDE 2017-04-01                   No     2020-02-01        One year   \n",
       "2   3668-QPYBK 2019-10-01  2019-12-01 00:00:00     2019-12-01  Month-to-month   \n",
       "3   7795-CFOCW 2016-05-01                   No     2020-02-01        One year   \n",
       "4   9237-HQITU 2019-09-01  2019-11-01 00:00:00     2019-11-01  Month-to-month   \n",
       "5   9305-CDSKC 2019-03-01  2019-11-01 00:00:00     2019-11-01  Month-to-month   \n",
       "6   1452-KIOVK 2018-04-01                   No     2020-02-01  Month-to-month   \n",
       "7   6713-OKOMC 2019-04-01                   No     2020-02-01  Month-to-month   \n",
       "8   7892-POOKP 2017-07-01  2019-11-01 00:00:00     2019-11-01  Month-to-month   \n",
       "9   6388-TABGU 2014-12-01                   No     2020-02-01        One year   \n",
       "10  9763-GRSKD 2019-01-01                   No     2020-02-01  Month-to-month   \n",
       "11  7469-LKBCI 2018-10-01                   No     2020-02-01        Two year   \n",
       "12  8091-TTVAX 2015-04-01                   No     2020-02-01        One year   \n",
       "13  0280-XJGEX 2015-09-01  2019-10-01 00:00:00     2019-10-01  Month-to-month   \n",
       "14  5129-JLPIS 2018-01-01                   No     2020-02-01  Month-to-month   \n",
       "15  3655-SNQYZ 2014-05-01                   No     2020-02-01        Two year   \n",
       "16  8191-XWSZG 2015-10-01                   No     2020-02-01        One year   \n",
       "17  9959-WOFKT 2014-03-01                   No     2020-02-01        Two year   \n",
       "18  4190-MFLUW 2019-01-01  2019-11-01 00:00:00     2019-11-01  Month-to-month   \n",
       "19  4183-MYFRB 2018-05-01                   No     2020-02-01  Month-to-month   \n",
       "\n",
       "   paperless_billing             payment_method  monthly_charges  \\\n",
       "0                Yes           Electronic check        29.850000   \n",
       "1                 No               Mailed check        56.950001   \n",
       "2                Yes               Mailed check        53.849998   \n",
       "3                 No  Bank transfer (automatic)        42.299999   \n",
       "4                Yes           Electronic check        70.699997   \n",
       "5                Yes           Electronic check        99.650002   \n",
       "6                Yes    Credit card (automatic)        89.099998   \n",
       "7                 No               Mailed check        29.750000   \n",
       "8                Yes           Electronic check       104.800003   \n",
       "9                 No  Bank transfer (automatic)        56.150002   \n",
       "10               Yes               Mailed check        49.950001   \n",
       "11                No    Credit card (automatic)        18.950001   \n",
       "12                No    Credit card (automatic)       100.349998   \n",
       "13               Yes  Bank transfer (automatic)       103.699997   \n",
       "14               Yes           Electronic check       105.500000   \n",
       "15                No    Credit card (automatic)       113.250000   \n",
       "16                No               Mailed check        20.650000   \n",
       "17                No  Bank transfer (automatic)       106.699997   \n",
       "18                No    Credit card (automatic)        55.200001   \n",
       "19               Yes           Electronic check        90.050003   \n",
       "\n",
       "    total_charges internet_service  ... partner dependents multiple_lines  \\\n",
       "0       29.850000                1  ...     Yes         No             No   \n",
       "1     1889.500000                1  ...      No         No             No   \n",
       "2      108.150002                1  ...      No         No             No   \n",
       "3     1840.750000                1  ...      No         No             No   \n",
       "4      151.649994                0  ...      No         No             No   \n",
       "5      820.500000                0  ...      No         No            Yes   \n",
       "6     1949.400024                0  ...      No        Yes            Yes   \n",
       "7      301.899994                1  ...      No         No             No   \n",
       "8     3046.050049                0  ...     Yes         No            Yes   \n",
       "9     3487.949951                1  ...      No        Yes             No   \n",
       "10     587.450012                1  ...     Yes        Yes             No   \n",
       "11     326.799988               No  ...      No         No             No   \n",
       "12    5681.100098                0  ...     Yes         No            Yes   \n",
       "13    5036.299805                0  ...      No         No            Yes   \n",
       "14    2686.050049                0  ...      No         No             No   \n",
       "15    7895.149902                0  ...     Yes        Yes            Yes   \n",
       "16    1022.950012               No  ...      No         No             No   \n",
       "17    7382.250000                0  ...      No        Yes            Yes   \n",
       "18     528.349976                1  ...     Yes        Yes             No   \n",
       "19    1862.900024                0  ...      No         No             No   \n",
       "\n",
       "    dayofweek      month  year  tenure  has_crcard exited service_count  \n",
       "0   Wednesday    January  2020       0           0      0             1  \n",
       "1    Saturday      April  2017       3           0      0             2  \n",
       "2     Tuesday    October  2019       0           0      1             2  \n",
       "3      Sunday        May  2016       4           0      0             3  \n",
       "4      Sunday  September  2019       0           0      1             0  \n",
       "5      Friday      March  2019       0           0      1             3  \n",
       "6      Sunday      April  2018       2           1      0             2  \n",
       "7      Monday      April  2019       1           0      0             1  \n",
       "8    Saturday       July  2017       2           0      1             4  \n",
       "9      Monday   December  2014       6           0      0             2  \n",
       "10    Tuesday    January  2019       1           0      0             1  \n",
       "11     Monday    October  2018       2           1      0             0  \n",
       "12  Wednesday      April  2015       5           1      0             3  \n",
       "13    Tuesday  September  2015       4           0      1             4  \n",
       "14     Monday    January  2018       2           0      0             5  \n",
       "15   Thursday        May  2014       6           1      0             6  \n",
       "16   Thursday    October  2015       5           0      0             0  \n",
       "17   Saturday      March  2014       6           0      0             4  \n",
       "18    Tuesday    January  2019       0           1      1             2  \n",
       "19    Tuesday        May  2018       2           0      0             3  \n",
       "\n",
       "[20 rows x 28 columns]"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head(20)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2d6aa2d0",
   "metadata": {},
   "source": [
    "## EDA "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "47424ed3",
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_snsbar(df, x, y, title):\n",
    "    xlabel = str(x.replace('_', ' ').capitalize())\n",
    "    ylabel = str(y.replace('_', ' ').capitalize())\n",
    "    # create grouped data\n",
    "    data = df.groupby([x])[y].count().sort_values(ascending=False).reset_index()\n",
    "    fig, ax=plt.subplots(figsize=(10,6))\n",
    "    ax = sns.barplot(x = x, y = y, data=data)\n",
    "    ax.set_title(title, fontdict={'size':12})\n",
    "    ax.set_ylabel(ylabel, fontsize = 10)\n",
    "    ax.set_xlabel(xlabel, fontsize = 10)\n",
    "    ax.set_xticklabels(data[x], rotation=90);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "5a5776a7",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# correlation matrix of features\n",
    "plt.figure(figsize=(10, 10))\n",
    "corrMatrix = df.corr()\n",
    "sns.color_palette(\"magma\", as_cmap=True)\n",
    "sns.heatmap(corrMatrix, annot=True, cmap='magma')\n",
    "plt.title('Correlation Matrix Plot')\n",
    "plt.show();"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "13d5288b",
   "metadata": {},
   "source": [
    "First I will explore a relationship between contract type and revenue "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "591d0f85",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>type</th>\n",
       "      <th>monthly_charges</th>\n",
       "      <th>total_charges</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Two year</td>\n",
       "      <td>103005.851562</td>\n",
       "      <td>6.283254e+06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>257294.156250</td>\n",
       "      <td>5.305862e+06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>One year</td>\n",
       "      <td>95816.601562</td>\n",
       "      <td>4.467054e+06</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "             type  monthly_charges  total_charges\n",
       "0        Two year    103005.851562   6.283254e+06\n",
       "1  Month-to-month    257294.156250   5.305862e+06\n",
       "2        One year     95816.601562   4.467054e+06"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# total monthly charges grouped by payment type\n",
    "(df.groupby('type', as_index=False)\n",
    "     .agg({'monthly_charges': 'sum', 'total_charges': 'sum'})\n",
    "     .sort_values(by='total_charges', ascending=False, ignore_index=True)\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d9e9c74a",
   "metadata": {},
   "source": [
    "here we can see that though month-to-month contracts bring in the most revenue monthly, two year contracts actually bring in the most revenue"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "8e79065e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>type</th>\n",
       "      <th>service_count</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>6013</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Two year</td>\n",
       "      <td>4654</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>One year</td>\n",
       "      <td>3686</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "             type  service_count\n",
       "0  Month-to-month           6013\n",
       "1        Two year           4654\n",
       "2        One year           3686"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# services count grouped by contract type\n",
    "(df.groupby('type', as_index=False)\n",
    "     .agg({'service_count': 'sum'})\n",
    "     .sort_values(by='service_count', ascending=False, ignore_index=True)\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ebe442e0",
   "metadata": {},
   "source": [
    "Checking if there is an emphasis on churn by contract type"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "77bcf864",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>type</th>\n",
       "      <th>exited</th>\n",
       "      <th>% exit percent</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Month-to-month</td>\n",
       "      <td>1655</td>\n",
       "      <td>55.02</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>One year</td>\n",
       "      <td>166</td>\n",
       "      <td>24.07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Two year</td>\n",
       "      <td>48</td>\n",
       "      <td>20.91</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "             type  exited % exit percent\n",
       "0  Month-to-month    1655          55.02\n",
       "1        One year     166          24.07\n",
       "2        Two year      48          20.91"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# effect of contract type on customer churn\n",
    "contract_type_percent = df.groupby(\n",
    "    'type', as_index=False).agg(\n",
    "    {'exited': 'sum'}).sort_values(\n",
    "    by='exited', ascending=False, ignore_index=True)\n",
    "contract_type_effect = (df['type'].value_counts() / df['type'].value_counts().sum() * 100).tolist()\n",
    "contract_type_percent['% exit percent'] = ['{:.2f}'.format(x) for x in contract_type_effect]\n",
    "contract_type_percent"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2d003a44",
   "metadata": {},
   "source": [
    "we can see there is a much higher percentage of churned accounts in the month to month contract type and the two year contract type has the lowest churn percentage. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "d0406447",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot of contract type on customer churn\n",
    "plot_snsbar(df, 'type', 'exited', 'Plot of contract type on customer churn')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7f693ae0",
   "metadata": {},
   "source": [
    "month to month contracts churn more often than the other two contracts\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "21bb8360",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot of service count on customer churn\n",
    "plot_snsbar(df, 'service_count', 'exited', 'Plot of service count on customer churn')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d29d29d6",
   "metadata": {},
   "source": [
    "customers with 5 or more services are less likely to churn\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "67182b5a",
   "metadata": {},
   "source": [
    "### What are the top services patronized?\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "64aea762",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "      <th>...</th>\n",
       "      <th>7033</th>\n",
       "      <th>7034</th>\n",
       "      <th>7035</th>\n",
       "      <th>7036</th>\n",
       "      <th>7037</th>\n",
       "      <th>7038</th>\n",
       "      <th>7039</th>\n",
       "      <th>7040</th>\n",
       "      <th>7041</th>\n",
       "      <th>7042</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>online_security</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>online_backup</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>device_protection</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>tech_support</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>streaming_tv</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>streaming_movies</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>exited</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>7 rows × 7043 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                   0     1     2     3     4     5     6     7     8     9     \\\n",
       "online_security       0     1     1     1     0     0     0     1     0     1   \n",
       "online_backup         1     0     1     0     0     0     1     0     0     1   \n",
       "device_protection     0     1     0     1     0     1     0     0     1     0   \n",
       "tech_support          0     0     0     1     0     0     0     0     1     0   \n",
       "streaming_tv          0     0     0     0     0     1     1     0     1     0   \n",
       "streaming_movies      0     0     0     0     0     1     0     0     1     0   \n",
       "exited                0     0     1     0     1     1     0     0     1     0   \n",
       "\n",
       "                   ...  7033  7034  7035  7036  7037  7038  7039  7040  7041  \\\n",
       "online_security    ...     0     1     0     0     0     1     0     1     0   \n",
       "online_backup      ...     0     1     0     1     0     0     1     0     0   \n",
       "device_protection  ...     0     1     0     1     0     1     1     0     0   \n",
       "tech_support       ...     0     0     0     1     0     1     0     0     0   \n",
       "streaming_tv       ...     0     1     1     1     0     1     1     0     0   \n",
       "streaming_movies   ...     0     0     0     1     0     1     1     0     0   \n",
       "exited             ...     0     1     0     0     0     0     0     0     1   \n",
       "\n",
       "                   7042  \n",
       "online_security       1  \n",
       "online_backup         0  \n",
       "device_protection     1  \n",
       "tech_support          1  \n",
       "streaming_tv          1  \n",
       "streaming_movies      1  \n",
       "exited                0  \n",
       "\n",
       "[7 rows x 7043 columns]"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# we create a copy of the dataframe to use for encoding\n",
    "df_encode = df.copy()\n",
    "\n",
    "# encoding services offered \n",
    "online_security = {'online_security':{'Yes': 1, 'No': 0}}\n",
    "online_backup = {'online_backup':{'Yes': 1, 'No': 0}}\n",
    "device_protection = {'device_protection':{'Yes': 1, 'No': 0}}\n",
    "tech_support = {'tech_support':{'Yes': 1, 'No': 0}}\n",
    "streaming_tv = {'streaming_tv':{'Yes': 1, 'No': 0}}\n",
    "streaming_movies = {'streaming_movies':{'Yes': 1, 'No': 0}}\n",
    "\n",
    "df_encode.replace(online_security, inplace =True)\n",
    "df_encode.replace(online_backup, inplace =True)\n",
    "df_encode.replace(device_protection, inplace =True)\n",
    "df_encode.replace(tech_support, inplace =True)\n",
    "df_encode.replace(streaming_tv, inplace =True)\n",
    "df_encode.replace(streaming_movies, inplace =True)\n",
    "\n",
    "services_data = df_encode[['online_security', 'online_backup', 'device_protection', 'tech_support', 'streaming_tv', 'streaming_movies', 'exited']]\n",
    "services_data = services_data.transpose()\n",
    "services_data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "974e783e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>services</th>\n",
       "      <th>count</th>\n",
       "      <th>% service offered</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>streaming_movies</td>\n",
       "      <td>2732</td>\n",
       "      <td>16.841327</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>streaming_tv</td>\n",
       "      <td>2707</td>\n",
       "      <td>16.687215</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>online_backup</td>\n",
       "      <td>2429</td>\n",
       "      <td>14.973493</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>device_protection</td>\n",
       "      <td>2422</td>\n",
       "      <td>14.930342</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>tech_support</td>\n",
       "      <td>2044</td>\n",
       "      <td>12.600173</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>online_security</td>\n",
       "      <td>2019</td>\n",
       "      <td>12.446061</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>exited</td>\n",
       "      <td>1869</td>\n",
       "      <td>11.521391</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            services  count  % service offered\n",
       "0   streaming_movies   2732          16.841327\n",
       "1       streaming_tv   2707          16.687215\n",
       "2      online_backup   2429          14.973493\n",
       "3  device_protection   2422          14.930342\n",
       "4       tech_support   2044          12.600173\n",
       "5    online_security   2019          12.446061\n",
       "6             exited   1869          11.521391"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# getting dataframe showing services and percentage count \n",
    "services_data['count'] = services_data.sum(axis=1)\n",
    "services_df = services_data.reset_index(inplace=False)\n",
    "services_df = services_df[['index', 'count']].rename(columns={'index': 'services'})\n",
    "services_df['% service offered']  = services_df['count'] / services_df['count'].sum() * 100\n",
    "services = services_df.copy()\n",
    "services.sort_values('% service offered', axis = 0, ascending = False, inplace = True, ignore_index=True)\n",
    "services"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "8ddd8801",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot of top 5 Interconnect service by count \n",
    "services_pie = services.head(5)\n",
    "(services_pie.set_index('services').plot(y='% service offered', kind='pie', \n",
    "                      title = 'Pie chart showing relative size of the five popular services', \n",
    "                      figsize=(8, 8), autopct='%1.1f%%', shadow=True)\n",
    ");"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3a02e376",
   "metadata": {},
   "source": [
    "The distribution of services patronized appears to be fairly even."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "41b1db94",
   "metadata": {},
   "source": [
    "### When do customers churn?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "8f3c3113",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAmQAAAGwCAYAAAAHVnkYAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuNCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8QVMy6AAAACXBIWXMAAAsTAAALEwEAmpwYAAA0eElEQVR4nO3deZhkVX3/8fcHUAEVWZwQZBEXXIhRIKNBTeKCGlAjmKBAXJCQoNG4ZhE1v7gm0cSomMVIgghKVEJEUTGKKLgkLiACKi5AQECWAdkEF9Dv74972qlpeqa7h6k+VTPv1/PU0/eee+vWt3tquj917rnnpqqQJElSPxv1LkCSJGlDZyCTJEnqzEAmSZLUmYFMkiSpMwOZJElSZwYySZKkzgxk0hySnJbkD5fotf44yZVJfphkm6V4zYVayp9De70fJrn3GrZflORxt+P4707yhrV9vibbUr9fpXXJQKYNVvvj/qMWAq5sf6zvsshj7JykkmyyljXcAXgL8ISquktVXbM2x1kXkrwmyXt7vT5A+xlc2Oq5XeEpyXOSfH7dVbf0enwPSR6d5NKlfE1JBjLpd6rqLsAewHLgL5f49bcFNgW+scSvK028tf2gM+2vrQ2TgUwCquoy4OPAg2ZvS7JRkr9McnGSq5Icm+RubfNn29frWk/bw+d4/p2SvC3J99vjba3tfsC3R57/6TmeO9MDd0iSS5Jcm+R5SR6a5Jwk1yX5p4XUOnKsg5N8L8nVSV7Vtu0NvBI4oH0fZ4+Ucc8kX0hyY5JPJrl7e86mSd6b5JpWx1eSbDvH93BIko+MrH83yX+OrF+SZLe2XEnum+Qw4BnAX7R6PjJyyN3a9359kg8k2XSO13wg8K/Aw9vzrxvZvFWSj7Xv50tJ7jPyvAckOSXJD5J8O8nTZx97ZN/Tkrwhyf/M1JhkmyTHJbmh/Tx2Htn/Ea3t+vb1ESPbnpPkwlbT/yV5xjzfw2gdWyc5ur23rk3yoZFjfn7WvpXkvm35iUm+2V7zsiR/luTODP8P7tFe84dJ7rG693A7zqOTXJrkL9p77vIk+7Xjf6f9LF85UsNGSQ5PckF77xyfZOu2beY9emiS7wG3+T/R9ts3ydfaz/mC9v6dsbr36216/jJyCjxDD/EJ7T19A/Cc9m/8+rmOJ61zVeXDxwb5AC4CHteWd2TopXp9Wz8N+MO2/AfA+cC9gbsAHwTe07btDBSwyRpe53XAF4FfApYB/zPyOmt8/sj2f2XoSXsC8GPgQ+142wNXAY9aRK3/BmwGPAT4CfDAtv01wHtnvf5pwAXA/dpzTgPe2LY9F/gIsDmwMfBrwBZzfA/3Bq5j+AB4D+Bi4NKRbdcCG7X1Au7blt8NvGGOf7Mvt+NsDZwHPG81P7vnAJ+f1fZu4BrgYcAmwHHA+9u2OwOXAIe0bbsDVwO7rub4p7Wf9X2AuwHfBL4DPK49/1jg6Lbv1u37fFbbdlBb36a97g3A/du+2wG/srrvYY46PgZ8ANgKuMPIe2Gu73/053s58JtteStgj7b86Jl/nwW+hx8N3Ar8VXv9PwJWAP8B3BX4FeBHwL3a/i9ux9oBuBPwTuB9s96jx7afy2ZzfL8PA64HHs/wntoeeMAC3q9zfV8XsfJ3wGuAW4D92nE3W9PxfPhY1w97yLSh+1Drefg8cDrwN3Ps8wzgLVV1YVX9EHgFcGAWfkrjGcDrquqqqloBvJbhD/NivL6qflxVnwRuYvgDdlUNPXufYwgPC631tVX1o6o6GzibIZitydFV9Z2q+hFwPLBba7+FIVDct6p+VlVnVtUNs59cw5iwG9vzfgv4BPD9JA8AHgV8rqp+voifxdur6vtV9QOGQLjbPPvPdmJVfbmqbmUIZDPPfzJwUVUdXVW3VtVZwH8BT1vDsY6uqguq6nqGnqULqupT7dj/ycp/lycB362q97Rjvw/4FvA7bfvPgQcl2ayqLq+qBZ3CTrIdsA9DKL22qm6pqtMX+HO4Bdg1yRbtuV9dw77zvYdvAf66qm4B3g/cHTiiqm5s38s3Wfk+ex7wqqq6tKp+whCE9p/1Hn1NVd3U3nOzHQq8q6pOqaqfV9VlVfWtke2re78uxP9W1YfacWde+/YcT1owA5k2dPtV1ZZVdc+qev5q/gDM9OrMuJihl+M2p+dWY67n32ORdV45svyjOdZnLkZYSK1XjCzfPPLc1Vnd/u9hCFfvb6ex/i7DRQpzOZ2hh+K32vJpDGHsUW19MRZb/0Kff0/g1zOcfr2uBfVnAL+8hmOt7b8LbX37qroJOIAhqFzeTqc+YIHfy47AD6rq2gXuP+r3gCcCFyc5PXOcbh8x33v4mqr6WVue+T+0up/FPYETR37G5wE/Y9X36CVrqGVHhl6r1bk974+5Xvf2vt+kBTGQSfP7PsMfkRk7MZyiuZLh9MraPP/766y6+V9rptb5LOR7Wbnz0Bvz2qraFXgEQw/Ts1ez+0wg+822fDrzB7JF1bMOnn8JcHoL6DOPu1TVH9/OOuC2/y4w/NtcBlBVn6iqxzOcrvwWw2llmP97uATYOsmWc2y7ieF0MgBJVgmWVfWVqtqX4TTkhxh6f1b3muvyPXwJsM+sn/Omrbf3F+XN8/z7rGH76sz+eWzMcPp11O19z0lrzUAmze99wEuT3CvDtBh/A3ygnZZawXC6abVzZ7Xn/2WSZW1A8F8B45peYk21zudKYOckC/q9kOQxSX61/WG7geG01epOPZ4OPIZhTNClDKdZ92Y45XnWGupZ0891PlcCOyS54wL3/yhwvyTPSnKH9nhoG1x/e53cjv37STZJcgCwK/DRJNu2Qep3ZhjT90NW/hzX+D1U1eUMp0r/JclWrebfapvPBn4lyW4ZLnx4zczzktwxw4UDd2unGW+Y9ZrbZOWFK7Bu38P/Cvx1knu2WpYl2XcRzz8KOCTJXu0Cge0X2KP4HWDTJE9qPbl/yTCGTZoIBjJpfu9iOD33WeD/GAbVvxCgqm4G/hr4QjsFs+ccz38DcAZwDnAu8NXWtqS1LsDMlY/XJFnTeKIZvwycwPDH/DyG0PWeuXasqu8wBI3PtfUbgAuBL4yc6prtKIYxTtelXTm4SJ9muFDjiiRXz7dzVd3IcNHEgQy9P1cAb2Id/NGuYX65JwN/ynBRwV8AT66qqxl+D7+sveYPGHoNZ3rlFvI9PIshDH+L4QKPl7TX/A7DYPxPAd9lGCc5+3kXtSsKn8dwepY2Hut9wIXtZ38P1u17+AjgJOCTSW5kGOD/6wt9clV9meHCi7cyDO4/ndv2Ps71vOuB5wP/ztAzeRPgfGuaGKmyh1aSJKkne8gkSZI6M5BJkiR1ZiCTJEnqzEAmSZLUmYFMkiSps6m+m/3d73732nnnnXuXIUmSNK8zzzzz6qqaPSExMOWBbOedd+aMM87oXYYkSdK8ksy+hdoveMpSkiSpMwOZJElSZwYySZKkzgxkkiRJnRnIJEmSOjOQSZIkdWYgkyRJ6sxAJkmS1JmBTJIkqTMDmSRJUmcGMkmSpM7GGsiSvDTJN5J8Pcn7kmya5F5JvpTk/CQfSHLHtu+d2vr5bfvO46xNkiRpUowtkCXZHngRsLyqHgRsDBwIvAl4a1XdF7gWOLQ95VDg2tb+1rafJEnSem+TJTj+ZkluATYHLgceC/x+234M8BrgHcC+bRngBOCfkqSqajEvuOId7739Va9jy/74mb1LkCRJE2xsPWRVdRnwZuB7DEHseuBM4LqqurXtdimwfVveHrikPffWtv8246pPkiRpUozzlOVWDL1e9wLuAdwZ2HsdHPewJGckOWPFihW393CSJEndjXNQ/+OA/6uqFVV1C/BB4JHAlklmTpXuAFzWli8DdgRo2+8GXDP7oFV1ZFUtr6rly5YtG2P5kiRJS2Ocgex7wJ5JNk8SYC/gm8BngP3bPgcDH27LJ7V12vZPL3b8mCRJ0jQa5xiyLzEMzv8qcG57rSOBlwMvS3I+wxixo9pTjgK2ae0vAw4fV22SJEmTZKxXWVbVq4FXz2q+EHjYHPv+GHjaOOuRJEmaRM7UL0mS1JmBTJIkqTMDmSRJUmcGMkmSpM4MZJIkSZ0ZyCRJkjozkEmSJHVmIJMkSerMQCZJktSZgUySJKkzA5kkSVJnBjJJkqTODGSSJEmdGcgkSZI6M5BJkiR1ZiCTJEnqzEAmSZLUmYFMkiSpMwOZJElSZwYySZKkzgxkkiRJnRnIJEmSOjOQSZIkdWYgkyRJ6sxAJkmS1JmBTJIkqTMDmSRJUmcGMkmSpM4MZJIkSZ0ZyCRJkjobWyBLcv8kXxt53JDkJUm2TnJKku+2r1u1/ZPk7UnOT3JOkj3GVZskSdIkGVsgq6pvV9VuVbUb8GvAzcCJwOHAqVW1C3BqWwfYB9ilPQ4D3jGu2iRJkibJUp2y3Au4oKouBvYFjmntxwD7teV9gWNr8EVgyyTbLVF9kiRJ3SxVIDsQeF9b3raqLm/LVwDbtuXtgUtGnnNpa5MkSVqvjT2QJbkj8BTgP2dvq6oCapHHOyzJGUnOWLFixTqqUpIkqZ+l6CHbB/hqVV3Z1q+cORXZvl7V2i8Ddhx53g6tbRVVdWRVLa+q5cuWLRtj2ZIkSUtjKQLZQaw8XQlwEnBwWz4Y+PBI+7Pb1ZZ7AtePnNqUJElab20yzoMnuTPweOC5I81vBI5PcihwMfD01n4y8ETgfIYrMg8ZZ22SJEmTYqyBrKpuAraZ1XYNw1WXs/ct4AXjrEeSJGkSOVO/JElSZwYySZKkzgxkkiRJnRnIJEmSOjOQSZIkdWYgkyRJ6sxAJkmS1JmBTJIkqTMDmSRJUmcGMkmSpM4MZJIkSZ0ZyCRJkjozkEmSJHVmIJMkSerMQCZJktSZgUySJKkzA5kkSVJnBjJJkqTODGSSJEmdGcgkSZI6M5BJkiR1ZiCTJEnqzEAmSZLUmYFMkiSps016F6DB996+f+8SbmOnF52woP0+cdQTx1zJ4v32oSf3LkGSpAUzkGmD9c73/HbvEm7juc/6RO8SJEkdeMpSkiSpMwOZJElSZwYySZKkzgxkkiRJnY01kCXZMskJSb6V5LwkD0+ydZJTkny3fd2q7Zskb09yfpJzkuwxztokSZImxbh7yI4A/ruqHgA8BDgPOBw4tap2AU5t6wD7ALu0x2HAO8ZcmyRJ0kQY27QXSe4G/BbwHICq+inw0yT7Ao9uux0DnAa8HNgXOLaqCvhi613brqouH1eN0rQ65MS9e5dwG0c/9b97lyBJU2uc85DdC1gBHJ3kIcCZwIuBbUdC1hXAtm15e+CSkedf2tpWCWRJDmPoQWOnnXYaW/GS1r0nnfj3vUu4jY899c97lyBJYz1luQmwB/COqtoduImVpycBaL1htZiDVtWRVbW8qpYvW7ZsnRUrSZLUyzgD2aXApVX1pbZ+AkNAuzLJdgDt61Vt+2XAjiPP36G1SZIkrdfGFsiq6grgkiT3b017Ad8ETgIObm0HAx9uyycBz25XW+4JXO/4MUmStCEY970sXwgcl+SOwIXAIQwh8PgkhwIXA09v+54MPBE4H7i57StJE+HJJxzXu4Tb+Oj+z+hdgqR1ZKyBrKq+BiyfY9Nec+xbwAvGWY8kSdIkcqZ+SZKkzgxkkiRJnRnIJEmSOjOQSZIkdWYgkyRJ6sxAJkmS1JmBTJIkqTMDmSRJUmcGMkmSpM4MZJIkSZ0ZyCRJkjozkEmSJHVmIJMkSerMQCZJktSZgUySJKkzA5kkSVJnBjJJkqTODGSSJEmdGcgkSZI6M5BJkiR1ZiCTJEnqbJPeBUiSxme/E07tXcJtfGj/vXqXIE0ce8gkSZI6M5BJkiR15ilLSdJEetGJl/Qu4Tbe/tQde5eg9ZQ9ZJIkSZ3ZQyZJ0jr08Q9c3buE29jngLv3LkHzsIdMkiSpMwOZJElSZ2M9ZZnkIuBG4GfArVW1PMnWwAeAnYGLgKdX1bVJAhwBPBG4GXhOVX11nPVJkqSVLnrbFb1LuI2dX/LLvUtYEkvRQ/aYqtqtqpa39cOBU6tqF+DUtg6wD7BLexwGvGMJapMkSequxynLfYFj2vIxwH4j7cfW4IvAlkm261CfJEnSkhp3ICvgk0nOTHJYa9u2qi5vy1cA27bl7YHRSWcubW2SJEnrtXFPe/EbVXVZkl8CTknyrdGNVVVJajEHbMHuMICddtpp3VUqSZLUyVh7yKrqsvb1KuBE4GHAlTOnItvXq9rulwGjUyDv0NpmH/PIqlpeVcuXLVs2zvIlSZKWxNh6yJLcGdioqm5sy08AXgecBBwMvLF9/XB7yknAnyR5P/DrwPUjpzYlSZLmdOUR/9u7hNvY9sUPX9T+4zxluS1w4jCbBZsA/1FV/53kK8DxSQ4FLgae3vY/mWHKi/MZpr04ZIy1SZIkTYyxBbKquhB4yBzt1wB7zdFewAvGVY8kSdKkcqZ+SZKkzgxkkiRJnRnIJEmSOjOQSZIkdWYgkyRJ6sxAJkmS1JmBTJIkqTMDmSRJUmcGMkmSpM4MZJIkSZ0ZyCRJkjozkEmSJHW2xpuLJ3nZmrZX1VvWbTmSJEkbnjUGMuCu7ev9gYcCJ7X13wG+PK6iJEmSNiRrDGRV9VqAJJ8F9qiqG9v6a4CPjb06SZKkDcBCx5BtC/x0ZP2nrU2SJEm303ynLGccC3w5yYltfT/gmLFUJEmStIFZUCCrqr9O8nHgN1vTIVV11vjKkiRJ2nAsZtqLzYEbquoI4NIk9xpTTZIkSRuUBQWyJK8GXg68ojXdAXjvuIqSJEnakCy0h+ypwFOAmwCq6vusnBJDkiRJt8NCA9lPq6qAAkhy5/GVJEmStGFZaCA7Psk7gS2T/BHwKeDfx1eWJEnShmOhV1m+OcnjgRsYZu3/q6o6ZayVSZIkbSAWFMiSvKmqXg6cMkebJEmSboeFnrJ8/Bxt+6zLQiRJkjZUa+whS/LHwPOBeyc5Z2TTXYEvjLMwSZKkDcV8pyz/A/g48LfA4SPtN1bVD8ZWlSRJ0gZkvkBWVXVRkhfM3pBka0OZJEnS7beQHrInA2cyzEGWkW0F3HtMdUmSJG0w1hjIqurJ7eta37cyycbAGcBlVfXkdg/M9wPbMAS9Z1XVT5PcCTgW+DXgGuCAqrpobV9XkiRpWiz0XpaHzlrfuN3fciFeDJw3sv4m4K1VdV/gWmDm2IcC17b2t7b9JEmS1nsLnfZiryQnJ9kuyYOAL7KAe1km2QF4Em1W/yQBHguc0HY5BtivLe/b1mnb92r7S5IkrdcWOlP/7yc5ADiX4Qbjv19VC5n24m3AX7AyvG0DXFdVt7b1S4Ht2/L2wCXt9W5Ncn3b/+qF1ChJkjStFnrKcheGU4//BVwMPCvJ5vM858nAVVV15u2uctXjHpbkjCRnrFixYl0eWpIkqYuFnrL8CPD/quq5wKOA7wJfmec5jwSekuQihkH8jwWOYLhB+UzP3A7AZW35MmBHgLb9bgyD+1dRVUdW1fKqWr5s2bIFli9JkjS5FhrIHlZVp8IwMVlV/QPw1DU9oapeUVU7VNXOwIHAp6vqGcBngP3bbgcDH27LJ7V12vZPV1Ut+DuRJEmaUmsMZEn+AqCqbkjytFmbn7OWr/ly4GVJzmcYI3ZUaz8K2Ka1v4xV7wwgSZK03pqvh+zAkeVXzNq290JfpKpOG5nT7MKqelhV3beqnlZVP2ntP27r923bL1zo8SVJkqbZfIEsq1mea12SJElrYb5AVqtZnmtdkiRJa2G+ecgekuQGht6wzdoybX3TsVYmSZK0gZjvXpYbL1UhkiRJG6qFTnshSZKkMTGQSZIkdWYgkyRJ6sxAJkmS1JmBTJIkqTMDmSRJUmcGMkmSpM4MZJIkSZ0ZyCRJkjozkEmSJHVmIJMkSerMQCZJktSZgUySJKkzA5kkSVJnBjJJkqTODGSSJEmdGcgkSZI6M5BJkiR1ZiCTJEnqzEAmSZLUmYFMkiSpMwOZJElSZwYySZKkzgxkkiRJnRnIJEmSOhtbIEuyaZIvJzk7yTeSvLa13yvJl5Kcn+QDSe7Y2u/U1s9v23ceV22SJEmTZJw9ZD8BHltVDwF2A/ZOsifwJuCtVXVf4Frg0Lb/ocC1rf2tbT9JkqT13tgCWQ1+2Fbv0B4FPBY4obUfA+zXlvdt67TteyXJuOqTJEmaFGMdQ5Zk4yRfA64CTgEuAK6rqlvbLpcC27fl7YFLANr264FtxlmfJEnSJBhrIKuqn1XVbsAOwMOAB9zeYyY5LMkZSc5YsWLF7T2cJElSd0tylWVVXQd8Bng4sGWSTdqmHYDL2vJlwI4AbfvdgGvmONaRVbW8qpYvW7Zs3KVLkiSN3TivslyWZMu2vBnweOA8hmC2f9vtYODDbfmktk7b/umqqnHVJ0mSNCk2mX+XtbYdcEySjRmC3/FV9dEk3wTen+QNwFnAUW3/o4D3JDkf+AFw4BhrkyRJmhhjC2RVdQ6w+xztFzKMJ5vd/mPgaeOqR5IkaVI5U78kSVJnBjJJkqTODGSSJEmdGcgkSZI6M5BJkiR1ZiCTJEnqzEAmSZLUmYFMkiSpMwOZJElSZwYySZKkzgxkkiRJnRnIJEmSOjOQSZIkdWYgkyRJ6sxAJkmS1JmBTJIkqTMDmSRJUmcGMkmSpM4MZJIkSZ0ZyCRJkjozkEmSJHVmIJMkSerMQCZJktSZgUySJKkzA5kkSVJnBjJJkqTODGSSJEmdGcgkSZI6M5BJkiR1ZiCTJEnqbGyBLMmOST6T5JtJvpHkxa196ySnJPlu+7pVa0+Styc5P8k5SfYYV22SJEmTZJw9ZLcCf1pVuwJ7Ai9IsitwOHBqVe0CnNrWAfYBdmmPw4B3jLE2SZKkiTG2QFZVl1fVV9vyjcB5wPbAvsAxbbdjgP3a8r7AsTX4IrBlku3GVZ8kSdKkWJIxZEl2BnYHvgRsW1WXt01XANu25e2BS0aedmlrkyRJWq+NPZAluQvwX8BLquqG0W1VVUAt8niHJTkjyRkrVqxYh5VKkiT1MdZAluQODGHsuKr6YGu+cuZUZPt6VWu/DNhx5Ok7tLZVVNWRVbW8qpYvW7ZsfMVLkiQtkXFeZRngKOC8qnrLyKaTgIPb8sHAh0fan92uttwTuH7k1KYkSdJ6a5MxHvuRwLOAc5N8rbW9EngjcHySQ4GLgae3bScDTwTOB24GDhljbZIkSRNjbIGsqj4PZDWb95pj/wJeMK56JEmSJpUz9UuSJHVmIJMkSerMQCZJktSZgUySJKkzA5kkSVJnBjJJkqTODGSSJEmdGcgkSZI6M5BJkiR1ZiCTJEnqzEAmSZLUmYFMkiSpMwOZJElSZwYySZKkzgxkkiRJnRnIJEmSOjOQSZIkdWYgkyRJ6sxAJkmS1JmBTJIkqTMDmSRJUmcGMkmSpM4MZJIkSZ0ZyCRJkjozkEmSJHVmIJMkSerMQCZJktSZgUySJKkzA5kkSVJnBjJJkqTOxhbIkrwryVVJvj7StnWSU5J8t33dqrUnyduTnJ/knCR7jKsuSZKkSTPOHrJ3A3vPajscOLWqdgFObesA+wC7tMdhwDvGWJckSdJEGVsgq6rPAj+Y1bwvcExbPgbYb6T92Bp8EdgyyXbjqk2SJGmSLPUYsm2r6vK2fAWwbVveHrhkZL9LW9ttJDksyRlJzlixYsX4KpUkSVoi3Qb1V1UBtRbPO7KqllfV8mXLlo2hMkmSpKW11IHsyplTke3rVa39MmDHkf12aG2SJEnrvaUOZCcBB7flg4EPj7Q/u11tuSdw/cipTUmSpPXaJuM6cJL3AY8G7p7kUuDVwBuB45McClwMPL3tfjLwROB84GbgkHHVJUmSNGnGFsiq6qDVbNprjn0LeMG4apEkSZpkztQvSZLUmYFMkiSpMwOZJElSZwYySZKkzgxkkiRJnRnIJEmSOjOQSZIkdWYgkyRJ6sxAJkmS1JmBTJIkqTMDmSRJUmcGMkmSpM4MZJIkSZ0ZyCRJkjozkEmSJHVmIJMkSerMQCZJktSZgUySJKkzA5kkSVJnBjJJkqTODGSSJEmdGcgkSZI6M5BJkiR1ZiCTJEnqzEAmSZLUmYFMkiSpMwOZJElSZwYySZKkzgxkkiRJnU1UIEuyd5JvJzk/yeG965EkSVoKExPIkmwM/DOwD7ArcFCSXftWJUmSNH4TE8iAhwHnV9WFVfVT4P3Avp1rkiRJGrtJCmTbA5eMrF/a2iRJktZrqareNQCQZH9g76r6w7b+LODXq+pPZu13GHBYW70/8O0xlXR34OoxHXvcprX2aa0bprf2aa0bprf2aa0bprf2aa0bprf2aa0bxlv7Patq2VwbNhnTC66Ny4AdR9Z3aG2rqKojgSPHXUySM6pq+bhfZxymtfZprRumt/ZprRumt/ZprRumt/ZprRumt/ZprRv61T5Jpyy/AuyS5F5J7ggcCJzUuSZJkqSxm5gesqq6NcmfAJ8ANgbeVVXf6FyWJEnS2E1MIAOoqpOBk3vX0Yz9tOgYTWvt01o3TG/t01o3TG/t01o3TG/t01o3TG/t01o3dKp9Ygb1S5IkbagmaQyZJEnSBslAJkmS1JmBbMoleWGSrXrXIUmTJMlmSe7fu47FSPIPSX6ldx3qw0A2/bYFvpLk+HZz9vQuaKGSbJzkuN51rI1W+5t717FY1q3FSHLqQtomTZLfAb4G/Hdb3y3JNEyjdB5wZJIvJXlekrv1Lmg+SbZe06N3fQuV5Fe71+Cg/pWSnAm8C/iPqrq2dz0L1ULYE4BDgOXA8cBRVXVB18IWIMnngce2+5dOlSRfrKo9e9exWNa99JIsA/4I2JmRq9ur6g961bQmSTYFNgc+AzwamPmgtwXw31X1gE6lLUj7Xf5Y4LSq2r21nVtV3f/oLkTr2TsEOAj4AvBvVfWZvlXNLcn/AcXK98ioqqp7L3FJayXJ54A7Ae8Gjquq65e6homa9mICHMDwn+ArSc4AjgY+WROeWquqklwBXAHcCmwFnJDklKr6i77VzetC4Avt0+tNM41V9ZZ+JS3YWa3u/2TV2j/Yr6QFse6l92Hgc8CngJ91rmUhngu8BLgHcCYr/9jeAPxTp5oW45aqun7WCYOJ/j0+I8nGwAPa42rgbOBlSZ5bVQd2LW4OVXWv3jWsC1X1m0l2Af4AODPJl4Gjq+qUparBHrI5JNkIeDLwDoZfnkcDR1TVD7oWNockLwaezfAf99+BD1XVLe17+G5V3adrgfNI8uq52qvqtUtdy2IlOXqO5prUXo8Z1r30knytqnbrXcdiJXlhVf1j7zoWK8lRwKnA4cDvAS8C7lBVz+ta2DySvBX4HYbaj6qqL49s+3ZVTdyYuCQPqKpvJdljru1V9dWlrun2aIF4P+DtDB9AArxyKT74GchmSfJghl6yJzLcNeA44DeAZ03iL9Qkr2FI8RfPse2BVXXe0le1eEk2r6qbe9chjUOSNwD/0ya/nhpJnsZwivLGJH8J7AG8YdL/yCbZHHgVw1COMPwuf31V/bhrYfNIcghwfFXdNMe2u/U4jTafJEdW1WFJ5jqlWlX12CUvai2M/O1/EnAKQyD+apJ7AP9bVfccew0GspXauIPrgKOA/6qqn4xs+2BV/W6v2ubSkvw3Jn08x5okeTjDz/suVbVTkocAz62q53cubV5J7sfQi7ptVT2o/Yd+SlW9oXNpa2TdSyfJjawcX3Nn4CfALW29qmqLjuXNK8k5VfXgJL8BvAH4e+CvqurXO5e2Xlld79KMKQjAGwEPr6ov9K5lbSU5neEs0wlV9aNZ255VVe8Zew0GskF7Qx1eVX/Tu5bFSPJh4IVV9b3etayNJF8C9gdOGhl8+/WqelDfyubX/gP/OfDOaardurVQSc6qqt2T/C1wblX9x0xb79rWpIX3P+O2F1FMZG/NanqXZkxFL9M0vC8mnYP6m6r6eZLfBaYqkDEM4P9GG4A4OtD5Kf1KWpyqumTW4NtpGPQMsHlVfXlW7bf2KmYRrHuJJXkq8OmZU05JtgQeXVUf6lnXAlyW5J3A44E3JbkT0zFd0n8C/8rQ4zHxv0+q6jG9a1gHTk3ye8AHJ/1CuLm0Af1/C+wKbDrTvpRXiRrIVvWpJH8GfIBVw83EDeYf8f96F3A7XZLkEUAluQPwYoa5eKbB1UnuQ7t6K8n+wOV9S1oQ6156r66qE2dWquq6dkHLh/qVtCBPB/YG3txq3o6hl3LS3VpV7+hdxGIlefZc7VV17FLXshaeC7wM+FmSHzElp+VHHA28Gngr8BiG8WRL+uHDU5Yj2nwqs03NPCrTKMndgSOAxzG8+T8BvLiqrula2AIkuTdwJPAI4Frg/4BnzHWBxSSx7qU3MxZrVtvEz4uVZKe52id9iES72Okq4ESGcXvAxH+4JsnoFa2bAnsBX62q/TuVtMFIcmZV/dro/8uZtiWrwUA23ZLsCfwj8EDgjsDGwE1T9Klk6iW5M7BRVd3Yu5bFsO6lk+RdDBcM/XNregGwdVU9p1dNC5HkXFZelLApcC/g21U10bf3WV8+XLdT2++vqr1717IQbdjPbzC8Zz43BafkfyHJ/zDUfgLwaeAy4I1LOdWIgWyWJA/itueQJ7a7uE1geyDDmInlDHOS3a+qXtG1sAVqvR5HAHsy/Cf+X+ClVXVh18IWIMk2DF3cM7+APg+8btJ796x76bUQ+f8YeoKL4bL6v55reoNJ1q4GfH5V/WHvWjYEbRjH1ydx/rHZkvwLcF/gfa3pAOCCqnpBv6oWLslDGYbLbAm8Hrgb8HdV9cUlq8FAtlIb0/FohkB2MrAP8PlJ7i5OckZVLR89JTJNV7sk+SJDr8HMf+IDGa4anfjL6pOcAnwWeG9regbDQO3H9atqfta9tNr0NJ9aTwZuT8up1qkci5XkI6y8o8DGDH+Ljq+ql/eramGSfAt44MyA/jZzwTeq6oF9K5seDupf1f7AQ4CzquqQJNuy8pf/pLo5yR2BryX5O4ZBztNwFdSMzWfN7/LeJNMwaBhgu6p6/cj6G5Ic0K2ahbPuJVRVP0vy80md2HNNkrxsZHUjholhv9+pnMV46MjyL8ZiARMdyIA3szKQ3QpcXFWXdaxnMc4HdgJmxnTu2NomWua56fxSzlhgIFvVj9r0F7cm2YJhUOiOvYuax7MYPkn9CfBShnp/r2tFC5Bk67b48SSHA+9n+EV0AEPv5DT4ZJIDGW7mDkOg/0THehbKupfeD4FzWy/f6BXcL+pX0oLcdWT5VuBjwH91qmXBquqFo+szY7H6VDO/WRMIj6okPwEuAF5VVacueXELd1fgvDYFEwyh+CszgWeCp2J6OHAJw1maLzH3TdKXhKcsR7Rz4K9kOG32pwy/RL9WVYd0LWw91AbdzvULCCZ88O0cs6//vG3aCPjhpF5QYd39JDl4rvaqOmapa9kQTdNYrNnaKe8HAcdN8iTISR41ugr8JsPf0ucDVNXpPeqaT/v5Ph44CHgww4eO91XVN5a8FgPZ3JLsDGxRVef0rmVNRoLNKiY50EiaDrPGNM24HjiD4Y4JE3lvyGkei7U6SZ5bVe/sXceaJNkd+H3gaQzT0nywpujm9G3i44MYbhH22qr6pyV9fQPZSkl+a672qvrsUteyUO3KsxmbMvxH2Lqq/qpTSYvSPp08idve4uQtvWpajAz3U9yZVWv/YLeCFmiK634KMPP/9LSq+mjPehZqEmYBXxtJjgCWseqVczcwhJ0tqupZvWpbk9ZbM61jsaZKu03VQe1xNcPE6n9WS3Az7nWlBbEnMXwPOwMnAe9a6veMgWxE+1Q1Y1PgYcCZNQX3ERu11JPZ3R5JTgZ+DJzLylNRVNVruxW1QG1uqQcD32Bl7VVVf9CvqvlNcd1vZBiXclxrOgg4YxqmeEnyeVbOAv47tFnAJ/2DU5KvVNVD52pL8o1Jm49sTWOxGCaInYaxWFMlyc+BzwGHVtX5re3CSf+wMSPJsQynhE9mmPPt691qMZCtXpIdgbdV1cQOkm/zAs3YiGEusj+uqod0KmlR5prBfFok+WZV7dq7jsWa4rrPAXarqp+39Y0Zroie+PfPJMwCvjaSnAf89szM/G3m/k9U1QOnaXodmJ6xWNMmyX4MY8UeCfw3w8UT/15V9+pZ10K1QDlzoc1oIFryWz95leWaXcowA/4k+4eR5VuBixjuPzctPp7kCVX1yd6FrIX/TbJrVX2zdyGLNK11wzBp48ztb+7WsY7F+kmbl+m7Sf6EYRbwu3SuaSH+FPh8kgsY/kDdC3h+m+h2qi5IqKqfAWdn1dsT6XZqs/F/qL0n9gVeAvxSkncAJ0767/aqmphpouwhG9H+o878QDYCdgMuqqpnditqPZfkqQxzvW0E3EKHTyVrq41TOQm4guF0yEztE91jM8V1HwS8EfgMQ82/BbyiqiZ2OoMZkzAL+Npq42se0Fa/PakD+TU5kmzFMJ75gKraq3c908JANmLWpem3MoSxL/SqZyHaL8vf47YDtF/Xq6bFaFeJ7gucW1P2ZkxyPvAybjv+baJvdj2tdQMk2Y6Vk35+uaqu6FnP+m5aZ7yXppGnLJs2vuAJVfWM3rUs0ocZLkM/k6G3Y9pcwjA/0FSFsWZFVa1xlucJNZV1Jzm1fdo+aY62iTRJs4CvpWmd8V6aOgaypt3e5J5J7lhVP+1dzyLsUFV79y7idrgQOC3JxxkJlFMy7cVZSf4D+Air1j7p00dMVd1JNgU2B+7eToXMXEG3BbB9t8IWZmJmAV8b0zbjvTTNDGSruhD4QvtUO3p7k0kOB/+T5Fer6tzehayl/2uPO7bHNNmMIdA8YaStgIkMNiOmre7nMgwUvgdDT/BMqLkBWNKJG9fCL7NyFvDfp+Ms4OvITcBUTGcgTRvHkI1I8uq52id5Tqwk3wTuyxBqpmaA9mxJNq+qm3vXocmV5IXTNOv3bL1nAV8bc8x4/0CGGe8P71eVtH6yh2zETPBqNxavqrqxc0kLsU/vAm6PJA8HjmKYAmCnJA8BnltVz+9b2fzaDNXvALatqge12e+fUlVv6FzaGk1r3cDPk2xZVdfBL67kOqiq/qVvWWs2xyzgbwdO7FnTIrx5ZPlWhlB2QKdapPXaxMy/MQmSLE9yLnAOcG6Ss5NM9MSNVXVxuzruRwyfZGce0+JtwG8D1wBU1dmsvDXOpPs34BUM03XQ7nt6YNeKFmZa6/6jmTAGUFXXAn/Ur5z5tVnA/xfYg6FX7KFV9fppuY1PDTeEvgF4MsMdEl7LMH2HpHXMHrJVvQt4flV9DiDJbwBHM9xmZiK1e/v9A8P4mquAezL8wpyoW5qsSVVdkqwy1vlnvWpZpM2r6suzar+1VzGLMK11b5wkM1fktiujJ33c4TMZxl29GHjRyM98oufbW839CVNVj+lamLQeM5Ct6mczYQygqj6fZNL/UL0e2BP4VFXtnuQxDH8EpsUlSR4BVJI7MPzhmpZP4FcnuQ+tRzLJ/sDlfUtakGmt+xPAB5K8s60/j+FWLRNrkmYBX6RvMdyf8Mkj9yd8ad+SpPWbg/pZ5X6Qz2a4Au19DH+sDgB+XFUv61XbfJKcUVXLk5wN7F5VP09y9hTdy/LuwBHA4xh6DT4JvKiqfrDGJ06AJPcGjgQeAVzLcGHFMyZ9gtXV1P3MqrqoZ13zSbIZwynKR7WmTwBHtVviaB2a9vsTStPIQAYk+cwaNldVPXbJilmkJJ8C9mO4pcw2DKctH1pVj+hZ10IleeTsuyHM1TaJkmzc5q+7M7DRlFwE8gvTUneSTYC/AQ5hmNMLYCeG4QSvrKpbetW2vhu5P+FBwGMZJoSd+PsTStPIQDblkmwO/Jihd+mZDJNlHjcNPUwASb5aVXvM1zaJknyPoffgA8Cnp+VuA0n+huE+ite19a2AP62qv+xa2GokeStwV+ClM+ExyV0Zxk7+qKpe3LO+DYX3J5TGy0A2Ism2DJ/E71FV+yTZFXh4VR3VubTbSHIjt72acmbE8I+BC4BXVdWpS1rYArXpLh7BMOHnW0c2bQE8dRpOubYw/GSGUzt7AB8F3l9Vn+9a2DySnFVVu89qm9gQnOS7wP1mB942qP9bVbVLn8okad2Z1gGn4/JuhnEp92jr32EIDBOnqu5aVVvMety1qu7KMDv4cxnGZk2qOzLMPbYJQ+/HzOMGYP+OdS1YVd1cVcdX1e8CuzOEydM7l7UQG7e5sYBfjM260xr2763m6n1sY8f8RClpveBVlqu6e1Udn+QVAFV1a5KpGzDc/lCdnWRiZzVv8xudnuTdVXXxlE3G+wtJHsVw8cfewBnA0/tWtCDHAacmObqtHwIc07Ge+XwzybOrapUbWid5JsPVgJI09Qxkq7opyTasnA5gT+D6viWtvap65/x7dbcsyUcZesdIcj3wB1V1Zt+y5pfkIuAs4Hjgz6vqpjU/YzJU1ZvaVbmPa02vr6pP9KxpHi8APpjkDxjuZQmwnOGK6Kd2q0qS1iHHkI1o01/8I/Ag4OvAMmD/NpO5xiDJOcALZk3G+y/TcC/OJFtU1Q2961gbSe4J7FJVn2pj4Tae9N7JJI9l5YTH35zU8ZGStDYMZECSnarqe215E+D+DAPkv+0l9eM1bQPMRyXZgSHAP7I1fQ54cVVd2q+q+SX5I+AwYOuquk+SXYB/9co5SerHQf2DD40sf6CqvlFVXzeMLYnTk7wzyaOTPCrJvwCnJdljZMLeSXU0cBLDRSD3AD7S2ibdCxhC5A0AVfVd4Je6ViRJGzjHkA1Gb+p3725VbJhmprd49az23RnG8k3spLzAsqoaDWDvTvKSXsUswk+q6qcz91VsvcJ2lUtSRwayQa1mWWM25TcrvqZd6fe+tn4QcE3Hehbq9CSvBDZL8njg+Qy9e5KkThxDBrSpLW5i6CnbDLh5ZhPDVAxb9KptfTdNk/HO1gbG/yPwcIYg/z8M9+H8XtfC5pFkI+BQ4AkM7/FPMNyn0F8GktSJgUxdJfk4w7irV1XVQ9rps7Oq6lc7l7ZeS7IMoKpW9K5FkuQpS/U3dZPxJvmrNWyuqnr9khWzCBkGjb0a+BPaBT3tZ/2PVfW6nrVJ0obOqyzV2zROxnvTHA8YTgO+vFdRC/BShqsrH1pVW1fV1sCvA49M8tK+pUnShs1Tlupq2ifjTXJX4MUMYex44B+q6qq+Vc0tyVnA46vq6lnty4BPzp4PTpK0dOwhUxdJHprkl6vqq8CjgFcCPwE+CUz0xKoASbZO8gbgHIZT/3tU1csnNYw1d5gdxuAX48ju0KEeSVJjIFMv7wR+2pYfAbwK+GfgWuDIXkUtRJK/B74C3Aj8alW9pqqu7VzWQvx0LbdJksbMU5bqIsnZVfWQtvzPwIqqek1b/1pV7daxvDVK8nOG3rxbWXXeuomeJmVkepfbbAI2rSp7ySSpE6+yVC8bJ9mkqm4F9mK4t+KMiX5fVtVU9ixX1ca9a5AkzW2i//BpvfY+hhnjrwZ+xHBjbpLcl8m/ylKSpHXKU5bqpk1xsR3DFX43tbb7AXdpg/0lSdogGMgkSZI6m8qxMJIkSesTA5kkSVJnBjJJ66UkleS9I+ubJFmR5KNrebwtkzx/ZP3Ra3ssSZrNQCZpfXUT8KAkm7X1xwOX3Y7jbQk8f76dJGltGMgkrc9OBp7Ulg9imG4F+MXtrz6U5JwkX0zy4Nb+miTvSnJakguTvKg95Y3AfZJ8rd2tAeAuSU5I8q0kxyXJUn1jktYvBjJJ67P3Awcm2RR4MPClkW2vBc6qqgcz3Ev12JFtDwB+G3gY8OokdwAOBy6oqt2q6s/bfrsDLwF2Be4NPHKM34uk9ZiBTNJ6q6rOAXZm6B07edbm3wDe0/b7NLBNkpnbXn2sqn7SbsZ+FbDtal7iy1V1aVX9HPhaey1JWjRn6pe0vjsJeDPwaGCbBT7nJyPLP2P1vysXup8krZE9ZJLWd+8CXltV585q/xzwDBiumASurqob1nCcG4G7jqNASfLTnKT1WlVdCrx9jk2vAd6V5BzgZuDgeY5zTZIvJPk68HHgY+u6VkkbLm+dJEmS1JmnLCVJkjozkEmSJHVmIJMkSerMQCZJktSZgUySJKkzA5kkSVJnBjJJkqTODGSSJEmd/X/bLtjTgbWSnwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot to determine months with the most churn\n",
    "plot_snsbar(df, 'month', 'exited', 'Plot of months with the most customer churn')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a0642f7b",
   "metadata": {},
   "source": [
    "I would recommend that based on this graph, indicating high churn months to be February, September and December, that Telecom start incentive campagns in the months prior to these high churn months in an effort to reduce churn. For example in November January telecom could begin offering a free additional service (like movie streaming, or 'buyers choice') for a period of time to ward off some churn."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ae076cab",
   "metadata": {},
   "source": [
    "### EDA Conclusions\n",
    "From the exploratory data analasis we were able to learn the following:\n",
    "- Most customers prefer the month-to- month plan, with a majority of 61% of the customer base chosing this option\n",
    "- Customers with a 2 year contract are bringing in more revenue than the other two contract options \n",
    "- Customers on a 2 year contract are churning less than the other contracts\n",
    "- customers having 5 or more services with telecom are less likely to churn\n",
    "Recommendations:\n",
    "- incentivize 2 year contract participation, one month free, get two services for the price of one for the first year etc.\n",
    "- boost multiple service use by using promotions \n",
    "- introduce bonuses and incentives before peak churn months "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "44c99a6a",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\">\n",
    "<b>Reviewer's comment</b>\n",
    "\n",
    "Good job on data preprocessing and EDA! Dataframes were merged correctly. Tenure is a great feature!\n",
    "\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ae995d98",
   "metadata": {},
   "source": [
    "## Modeling "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "21b24793",
   "metadata": {},
   "source": [
    "In order to design the best model to predict customer churn I will train several models on the data and see which scores the highest on the AUC-ROC metric as well as check their accuracy. \n",
    "I will test:\n",
    "- dummy classifier\n",
    "- logistic regression\n",
    "- random forest classifier\n",
    "- Cat Boost\n",
    "- LightGBM"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7cae63ca",
   "metadata": {},
   "source": [
    "#### Splitting the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "666bb07e",
   "metadata": {},
   "outputs": [],
   "source": [
    "# drop unimportant features\n",
    "df = df.drop(['customer_id', 'begin_date', 'end_date', 'end_date_value', 'dayofweek', 'month', 'year'], axis=1)\n",
    "\n",
    "# declare variables for target and features\n",
    "y = df.exited\n",
    "X = df.drop(['exited'], axis=1)\n",
    "\n",
    "# split data into 75% training and 25% testing sets\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25, random_state=12345)\n",
    "\n",
    "X_train, X_valid, y_train, y_valid = train_test_split(X_train,y_train, test_size = 0.25, random_state=12345)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1e4e6666",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\">\n",
    "<b>Reviewer's comment</b>\n",
    "\n",
    "Alright, you dropped both dates and their components, so there should be no data leakage.\n",
    "    \n",
    "The data was split into train, validation and test sets.\n",
    "\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "2617ae2d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Categorical variables:\n",
      "['type', 'paperless_billing', 'payment_method', 'internet_service', 'online_security', 'online_backup', 'device_protection', 'tech_support', 'streaming_tv', 'streaming_movies', 'partner', 'dependents', 'multiple_lines']\n"
     ]
    }
   ],
   "source": [
    "# create copy of initial split feature dataset\n",
    "features_train = X_train.copy()\n",
    "features_test = X_test.copy()\n",
    "features_valid = X_valid.copy()\n",
    "\n",
    "# select numerical columns\n",
    "numerical_cols = [cname for cname in X_train.columns if X_train[cname].dtype in ['int64', 'int32', 'float32']]\n",
    "\n",
    "# list of categorical variables\n",
    "s = (features_train.dtypes == 'object')\n",
    "object_cols = list(s[s].index)\n",
    "print('Categorical variables:')\n",
    "print(object_cols)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "def783a2",
   "metadata": {},
   "source": [
    "#### One Hot Encoding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "70fd8d55",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Encoding features for machine learning\n",
    "\n",
    "# Approach 1: Ordinal Encoding\n",
    "# make a copy to avoid changing original data\n",
    "label__X_train = features_train.copy()\n",
    "label__X_test = features_test.copy()\n",
    "label__X_valid = features_valid.copy()\n",
    "\n",
    "features_train.replace(('Yes', 'No'), (1, 0), inplace=True)\n",
    "features_test.replace(('Yes', 'No'), (1, 0), inplace=True)\n",
    "features_valid.replace(('Yes', 'No'), (1, 0), inplace=True)\n",
    "\n",
    "# apply ordinal encoder to each column with categorical data\n",
    "ordinal_encoder = OrdinalEncoder()\n",
    "label__X_train[object_cols] = ordinal_encoder.fit_transform(features_train[object_cols])\n",
    "label__X_test[object_cols] = ordinal_encoder.transform(features_test[object_cols])\n",
    "label__X_valid[object_cols] = ordinal_encoder.transform(features_valid[object_cols])\n",
    "# Approach 2: One-Hot Encoding\n",
    "# one-hot encoding of categorical features\n",
    "df_ohe = pd.get_dummies(df, drop_first=True)\n",
    "\n",
    "# declare variables for target and features\n",
    "y_ohe = df_ohe.exited\n",
    "X_ohe = df_ohe.drop(['exited'], axis=1)\n",
    "\n",
    "# split data into 75% training and 25% testing sets\n",
    "X_train_ohe, X_test_ohe, y_train_ohe, y_test_ohe = train_test_split(X_ohe, y_ohe, test_size=0.25, random_state=12345)\n",
    "X_train_ohe, X_valid_ohe, y_train_ohe, y_valid_ohe = train_test_split(X_train_ohe, y_train_ohe, test_size=0.25, random_state=12345)\n",
    "# numerical features\n",
    "numerical_cols = [cname for cname in X_train_ohe.columns if X_train_ohe[cname].dtype in ['float32', 'float64', 'int64', 'int32']]\n",
    "\n",
    "# features scaling\n",
    "scaler = StandardScaler()\n",
    "scaler.fit(X_train_ohe[numerical_cols])\n",
    "# transform the training set and the test set using transform()\n",
    "X_train_ohe[numerical_cols] = scaler.transform(X_train_ohe[numerical_cols])\n",
    "X_test_ohe[numerical_cols]  = scaler.transform(X_test_ohe[numerical_cols])\n",
    "X_valid_ohe[numerical_cols]  = scaler.transform(X_valid_ohe[numerical_cols])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5e2dd4a9",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\">\n",
    "<b>Reviewer's comment</b>\n",
    "\n",
    "Great, you created two variants of the data with different encoding of categorical features: ordinal encoding and OHE. Note that you could use [OneHotEncoder](https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.OneHotEncoder.html) similarly to `OrdinalEncoder` instead of `pd.get_dummies()` to make your code a bit more streamlined.\n",
    "\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "771ab243",
   "metadata": {},
   "outputs": [],
   "source": [
    "# function to calculate model evaluation metrics\n",
    "def print_model_evaluation(y_valid, test_predictions):\n",
    "    print(\"\\033[1m\" + 'F1 score: ' + \"\\033[0m\", '{:.3f}'.format(f1_score(y_valid, test_predictions)))\n",
    "    print(\"\\033[1m\" + 'Accuracy Score: ' + \"\\033[0m\", '{:.2%}'.format(accuracy_score(y_valid, test_predictions)))\n",
    "    print(\"\\033[1m\" + 'Precision: ' + \"\\033[0m\", '{:.3f}'.format(precision_score(y_valid, test_predictions)))\n",
    "    print(\"\\033[1m\" + 'Recall: ' + \"\\033[0m\", '{:.3f}'.format(recall_score(y_valid, test_predictions)))\n",
    "    print(\"\\033[1m\" + 'Balanced Accuracy Score: ' + \"\\033[0m\", '{:.2%}'.format(balanced_accuracy_score(y_valid, test_predictions)))\n",
    "    print(\"\\033[1m\" + 'AUC-ROC Score: ' + \"\\033[0m\", '{:.2%}'.format(roc_auc_score(y_valid, test_predictions)))\n",
    "    print()\n",
    "    print(\"\\033[1m\" + 'Confusion Matrix' + \"\\033[0m\")\n",
    "    print('-'*50)\n",
    "    print(confusion_matrix(y_valid, test_predictions))\n",
    "    print()\n",
    "    print(\"\\033[1m\" + 'Classification report' + \"\\033[0m\")\n",
    "    print('-'*50)\n",
    "    print(classification_report(y_valid, test_predictions))\n",
    "    print()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cead4895",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-warning\">\n",
    "<b>Reviewer's comment</b>\n",
    "\n",
    "In `print_model_evaluation` function ROC-AUC is underestimated, because you're using binary predictions (method `predict`). To get the real ROC-AUC value you need to use probabilities (method `predict_proba`). The reason is that the ROC curve is constructed by varying the threshold of assigning positive class between 0 and 1 and for binary predictions the threshold is predefined.\n",
    "\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "id": "696fe214",
   "metadata": {},
   "outputs": [],
   "source": [
    "# function to evaluate classification models\n",
    "def evaluate_model(model, train_features, train_target, test_features, test_target):\n",
    "    \n",
    "    eval_stats = {}\n",
    "    \n",
    "    fig, axs = plt.subplots(1, 3, figsize=(20, 6)) \n",
    "    \n",
    "    for type, features, target in (('train', train_features, train_target), ('test', test_features, test_target)):\n",
    "        \n",
    "        eval_stats[type] = {}\n",
    "    \n",
    "        pred_target = model.predict(features)\n",
    "        pred_proba = model.predict_proba(features)[:, 1]\n",
    "        \n",
    "        # F1\n",
    "        f1_thresholds = np.arange(0, 1.01, 0.05)\n",
    "        f1_scores = [metrics.f1_score(target, pred_proba>=threshold) for threshold in f1_thresholds]\n",
    "        \n",
    "        # ROC\n",
    "        fpr, tpr, roc_thresholds = metrics.roc_curve(target, pred_proba)\n",
    "        roc_auc = metrics.roc_auc_score(target, pred_proba)    \n",
    "        eval_stats[type]['ROC AUC'] = roc_auc\n",
    "\n",
    "        # PRC\n",
    "        precision, recall, pr_thresholds = metrics.precision_recall_curve(target, pred_proba)\n",
    "        aps = metrics.average_precision_score(target, pred_proba)\n",
    "        eval_stats[type]['APS'] = aps\n",
    "        \n",
    "        if type == 'train':\n",
    "            color = 'blue'\n",
    "        else:\n",
    "            color = 'green'\n",
    "         # F1 Score\n",
    "        ax = axs[0]\n",
    "        max_f1_score_idx = np.argmax(f1_scores)\n",
    "        ax.plot(f1_thresholds, f1_scores, color=color, label=f'{type}, max={f1_scores[max_f1_score_idx]:.2f} @ {f1_thresholds[max_f1_score_idx]:.2f}')\n",
    "        # setting crosses for some thresholds\n",
    "        for threshold in (0.2, 0.4, 0.5, 0.6, 0.8):\n",
    "            closest_value_idx = np.argmin(np.abs(f1_thresholds-threshold))\n",
    "            marker_color = 'orange' if threshold != 0.5 else 'red'\n",
    "            ax.plot(f1_thresholds[closest_value_idx], f1_scores[closest_value_idx], color=marker_color, marker='X', markersize=7)\n",
    "        ax.set_xlim([-0.02, 1.02])    \n",
    "        ax.set_ylim([-0.02, 1.02])\n",
    "        ax.set_xlabel('threshold')\n",
    "        ax.set_ylabel('F1')\n",
    "        ax.legend(loc='lower center')\n",
    "        ax.set_title(f'F1 Score') \n",
    "\n",
    "        # ROC\n",
    "        ax = axs[1]    \n",
    "        ax.plot(fpr, tpr, color=color, label=f'{type}, ROC AUC={roc_auc:.2f}')\n",
    "        # setting crosses for some thresholds\n",
    "        for threshold in (0.2, 0.4, 0.5, 0.6, 0.8):\n",
    "            closest_value_idx = np.argmin(np.abs(roc_thresholds-threshold))\n",
    "            marker_color = 'orange' if threshold != 0.5 else 'red'            \n",
    "            ax.plot(fpr[closest_value_idx], tpr[closest_value_idx], color=marker_color, marker='X', markersize=7)\n",
    "        ax.plot([0, 1], [0, 1], color='grey', linestyle='--')\n",
    "        ax.set_xlim([-0.02, 1.02])    \n",
    "        ax.set_ylim([-0.02, 1.02])\n",
    "        ax.set_xlabel('FPR')\n",
    "        ax.set_ylabel('TPR')\n",
    "        ax.legend(loc='lower center')        \n",
    "        ax.set_title(f'ROC Curve')\n",
    "        \n",
    "        # PRC\n",
    "        ax = axs[2]\n",
    "        ax.plot(recall, precision, color=color, label=f'{type}, AP={aps:.2f}')\n",
    "        # setting crosses for some thresholds\n",
    "        for threshold in (0.2, 0.4, 0.5, 0.6, 0.8):\n",
    "            closest_value_idx = np.argmin(np.abs(pr_thresholds-threshold))\n",
    "            marker_color = 'orange' if threshold != 0.5 else 'red'\n",
    "            ax.plot(recall[closest_value_idx], precision[closest_value_idx], color=marker_color, marker='X', markersize=7)\n",
    "        ax.set_xlim([-0.02, 1.02])    \n",
    "        ax.set_ylim([-0.02, 1.02])\n",
    "        ax.set_xlabel('recall')\n",
    "        ax.set_ylabel('precision')\n",
    "        ax.legend(loc='lower center')\n",
    "        ax.set_title(f'PRC')        \n",
    "\n",
    "        eval_stats[type]['Accuracy'] = metrics.accuracy_score(target, pred_target)\n",
    "        eval_stats[type]['F1'] = metrics.f1_score(target, pred_target)\n",
    "    \n",
    "    df_eval_stats = pd.DataFrame(eval_stats)\n",
    "    df_eval_stats = df_eval_stats.round(2)\n",
    "    df_eval_stats = df_eval_stats.reindex(index=('Accuracy', 'F1', 'APS', 'ROC AUC'))\n",
    "    \n",
    "    print(df_eval_stats)\n",
    "    \n",
    "    return   "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2c5673a3",
   "metadata": {},
   "source": [
    "### Dummy Classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "id": "9daacd21",
   "metadata": {},
   "outputs": [],
   "source": [
    "# baseline model using a dummy classifier\n",
    "dummy_clf = DummyClassifier(strategy=\"most_frequent\")\n",
    "dummy_clf.fit(features_train, y_train)\n",
    "dummy_clf_test_predictions = dummy_clf.predict(features_valid)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "id": "5f093a92",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1mF1 score: \u001b[0m 0.000\n",
      "\u001b[1mAccuracy Score: \u001b[0m 72.29%\n",
      "\u001b[1mPrecision: \u001b[0m 0.000\n",
      "\u001b[1mRecall: \u001b[0m 0.000\n",
      "\u001b[1mBalanced Accuracy Score: \u001b[0m 50.00%\n",
      "\u001b[1mAUC-ROC Score: \u001b[0m 50.00%\n",
      "\n",
      "\u001b[1mConfusion Matrix\u001b[0m\n",
      "--------------------------------------------------\n",
      "[[955   0]\n",
      " [366   0]]\n",
      "\n",
      "\u001b[1mClassification report\u001b[0m\n",
      "--------------------------------------------------\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.72      1.00      0.84       955\n",
      "           1       0.00      0.00      0.00       366\n",
      "\n",
      "    accuracy                           0.72      1321\n",
      "   macro avg       0.36      0.50      0.42      1321\n",
      "weighted avg       0.52      0.72      0.61      1321\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# evaluate baseline model\n",
    "print_model_evaluation(y_valid, dummy_clf_test_predictions)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "62dd338d",
   "metadata": {},
   "source": [
    "the table shows that the dummy classifier achieved a AUC-ROC of 50% and an accuracy of 50% now we will train other models to beat these metrics"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d6f33bd8",
   "metadata": {},
   "source": [
    "#### Class Imbalance Check"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "id": "eb2edc1a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0    0.778955\n",
      "1    0.221045\n",
      "dtype: float64\n",
      "\n",
      "Accuracy score before upsampling: 0.844\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# sanity check\n",
    "model = LogisticRegression(random_state=12345, solver='liblinear')\n",
    "model.fit(label__X_train, y_train) # train the model \n",
    "test_predictions = pd.Series(model.predict(label__X_valid))\n",
    "class_frequency = test_predictions.value_counts(normalize=True)\n",
    "print(class_frequency)\n",
    "class_frequency.plot(kind='bar');\n",
    "print()\n",
    "print('Accuracy score before upsampling: {:.3f}'.format(accuracy_score(y_valid, test_predictions)))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "afde758f",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-warning\">\n",
    "<b>Reviewer's comment</b>\n",
    "\n",
    "To check for class imbalance you need to look at train set targets, not predictions of a model.\n",
    "\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e8e780c2",
   "metadata": {},
   "source": [
    "As we can see we have a considerable class imbalance betweeen our churned and current customer base. I will use upsampling to balance the classes "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "id": "02138984",
   "metadata": {},
   "outputs": [],
   "source": [
    " def upsample(features, target, repeat):\n",
    "    features_zeros = features[target == 0]\n",
    "    features_ones = features[target == 1]\n",
    "    target_zeros = target[target == 0]\n",
    "    target_ones = target[target == 1]\n",
    "\n",
    "    features_upsampled = pd.concat([features_zeros] + [features_ones] * repeat)\n",
    "    target_upsampled = pd.concat([target_zeros] + [target_ones] * repeat)\n",
    "\n",
    "    features_upsampled, target_upsampled = shuffle(\n",
    "        features_upsampled, target_upsampled, random_state=12345\n",
    "    )\n",
    "    return features_upsampled, target_upsampled\n",
    "\n",
    "# new training set created\n",
    "features_upsampled, target_upsampled = upsample(\n",
    "    label__X_train, y_train, 5\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ba55feee",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\">\n",
    "<b>Reviewer's comment</b>\n",
    "\n",
    "It's nice that you tried upsampling!\n",
    "\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "48fd881c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0    0.521575\n",
      "1    0.478425\n",
      "dtype: float64\n",
      "\n",
      "Accuracy score after upsampling: 0.735\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# sanity check after upsampling\n",
    "model = LogisticRegression(random_state=12345, solver='liblinear')\n",
    "model.fit(features_upsampled, target_upsampled) # train the model \n",
    "test_predictions = pd.Series(model.predict(label__X_valid))\n",
    "class_frequency = test_predictions.value_counts(normalize=True)\n",
    "print(class_frequency)\n",
    "class_frequency.plot(kind='bar');\n",
    "print()\n",
    "print('Accuracy score after upsampling: {:.3f}'.format(accuracy_score(y_valid, test_predictions)))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "43ba1016",
   "metadata": {},
   "source": [
    "our accuracy went down but now our classes are much more balanced. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ffe0a775",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-warning\">\n",
    "<b>Reviewer's comment</b>\n",
    "\n",
    "Predictions of the model need not be balanced. Upsampling is applied to help the model better learn about the underrepresented class during training, but if the real data distribution is not balanced, then making balanced predictions will probably produce poor results. So, we really need to look at the metrics (not accuracy, as it poorly reflects performance on imbalanced datasets), not how balanced predictions are.\n",
    "\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "id": "14addb64",
   "metadata": {},
   "outputs": [],
   "source": [
    "# upsampling the other label encoded trainsets\n",
    "label__X_train_upsampled, y_train_upsampled = upsample(\n",
    "    label__X_train, y_train, 5\n",
    ")\n",
    "\n",
    "X_train_ohe_upsampled, y_train_ohe_upsampled = upsample(\n",
    "    X_train_ohe, y_train_ohe, 5\n",
    ")\n",
    "\n",
    "X_train_upsampled, y_train_upsampled = upsample(\n",
    "    X_train, y_train, 5\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8ab2bff7",
   "metadata": {},
   "source": [
    "### Logistic Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "id": "d63f9bd4",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The best hyperparameters are: {'C': 1, 'penalty': 'l2', 'solver': 'lbfgs'}\n",
      "CPU times: user 1.63 s, sys: 4.21 s, total: 5.84 s\n",
      "Wall time: 5.8 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    " \n",
    "grid_logreg = {\n",
    "    \"C\": [0.01, 0.1, 1],\n",
    "    'penalty': ['l2'], \n",
    "    'solver': ['lbfgs']\n",
    "}\n",
    "\n",
    "# define the model\n",
    "logreg_clf = LogisticRegression(random_state = 12345)\n",
    "\n",
    "grid_search_logreg = GridSearchCV(\n",
    "    estimator = logreg_clf,\n",
    "    param_grid = grid_logreg,\n",
    "    scoring = 'roc_auc',\n",
    "    cv = 10\n",
    ")\n",
    "\n",
    "logreg_cv = grid_search_logreg.fit(X_train_ohe, y_train_ohe)\n",
    "\n",
    "print('The best hyperparameters are: {}'.format(logreg_cv.best_params_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "id": "2dbc4473",
   "metadata": {},
   "outputs": [],
   "source": [
    "# function to train model and make predictions\n",
    "def train_logistic_regression_classifier(X_train, y_train):\n",
    "    \"\"\"This function trains a logistic regression model\"\"\"\n",
    "    global logreg_model\n",
    "   \n",
    "    logreg_model = LogisticRegression(**grid_search_logreg.best_params_)\n",
    "    logreg_model.fit(X_train, y_train) \n",
    "    \n",
    "def logistic_regression_classifier_prediction(X_valid, y_valid):\n",
    "    \"\"\"\n",
    "    This function is used to make prediction \n",
    "    using a logistic regression classification model\n",
    "    \"\"\"\n",
    "    global logreg_pred, logreg_pred_proba\n",
    "    logreg_pred = logreg_model.predict(X_valid)\n",
    "    logreg_pred_proba = logreg_model.predict_proba(X_valid)[:, 1]\n",
    "    # AUC-ROC score\n",
    "    print(\"\\033[1m\" + 'AUC-ROC Score and Accuracy using Logistic Regression Classifier : ' + \"\\033[0m\")\n",
    "    print('AUC-ROC Score: {:.2%}'.format(roc_auc_score(y_valid, logreg_pred_proba)))\n",
    "    # accuracy for logistic regression classifier\n",
    "    logreg_accuracy = accuracy_score(y_valid, logreg_pred)\n",
    "    print('Accuracy score: {:.2%}'.format(logreg_accuracy))\n",
    "  "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1f0b5f05",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-warning\">\n",
    "<b>Reviewer's comment</b>\n",
    "\n",
    "Do we really need to use global variables here?\n",
    "\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "id": "61fe10c8",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 97.5 ms, sys: 172 ms, total: 269 ms\n",
      "Wall time: 293 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "train_logistic_regression_classifier(X_train_ohe, y_train_ohe)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "id": "14d2c164",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1mAUC-ROC Score and Accuracy using Logistic Regression Classifier : \u001b[0m\n",
      "AUC-ROC Score: 88.85%\n",
      "Accuracy score: 84.63%\n",
      "CPU times: user 5.3 ms, sys: 30.5 ms, total: 35.8 ms\n",
      "Wall time: 9.58 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "logistic_regression_classifier_prediction(X_valid_ohe, y_valid_ohe)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7832fc88",
   "metadata": {},
   "source": [
    "Logistic regression is able to score an AUC-ROC of 88%, lets see if we can do better with hyper parameter tuning in other models "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6ef04aea",
   "metadata": {},
   "source": [
    "### Random Forest Classifier with hyperparameter tuning "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "id": "00a9e987",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The best hyperparameters are: {'criterion': 'gini', 'max_depth': None, 'min_samples_leaf': 4, 'n_estimators': 50}\n",
      "CPU times: user 2min 10s, sys: 464 ms, total: 2min 11s\n",
      "Wall time: 2min 11s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "grid_rf = {\n",
    "    \"criterion\" : [\"gini\", \"entropy\"],\n",
    "    \"n_estimators\" : [10, 25, 50, 100],\n",
    "    \"max_depth\" : [None, 2, 4, 8, 10, 12],\n",
    "    \"min_samples_leaf\" : [2, 4, 6]\n",
    "}\n",
    " \n",
    "rf_clf = RandomForestClassifier(random_state = 12345)\n",
    "\n",
    "grid_search_rf = GridSearchCV(estimator=rf_clf, param_grid = grid_rf, cv=5, scoring = 'roc_auc')\n",
    "\n",
    "rf_cv = grid_search_rf.fit(label__X_train, y_train)\n",
    "\n",
    "print('The best hyperparameters are: {}'.format(rf_cv.best_params_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "id": "835697d0",
   "metadata": {},
   "outputs": [],
   "source": [
    "def train_random_forest_classifier(X_train, y_train):\n",
    "    \"\"\"This function trains a random forest model\"\"\"\n",
    "    global rf_model\n",
    "    rf_model = RandomForestClassifier(**grid_search_rf.best_params_)\n",
    "    rf_model.fit(X_train, y_train) \n",
    "    \n",
    "def random_forest_classifier_prediction(X_valid, y_valid):\n",
    "    \"\"\"\n",
    "    This function is used to make prediction \n",
    "    using a random forest classification model\n",
    "    \"\"\"\n",
    "    global rf_pred, rf_pred_proba\n",
    "    rf_pred = rf_model.predict(X_valid)\n",
    "    rf_pred_proba = rf_model.predict_proba(X_valid)[:, 1]\n",
    "    \n",
    "    print(\"\\033[1m\" + 'AUC-ROC Score and Accuracy using Random Forest Classifier' + \"\\033[0m\")\n",
    "    print('AUC-ROC Score: {:.2%}'.format(roc_auc_score(y_valid, rf_pred_proba)))\n",
    "    \n",
    "    rf_accuracy = accuracy_score(y_valid, rf_pred)\n",
    "    print('Accuracy score: {:.2%}'.format(rf_accuracy))\n",
    "   "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "id": "64031b18",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 221 ms, sys: 11.9 ms, total: 233 ms\n",
      "Wall time: 251 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "train_random_forest_classifier(label__X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "id": "2ed589ec",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1mAUC-ROC Score and Accuracy using Random Forest Classifier\u001b[0m\n",
      "AUC-ROC Score: 88.33%\n",
      "Accuracy score: 84.33%\n",
      "CPU times: user 39.2 ms, sys: 3.96 ms, total: 43.2 ms\n",
      "Wall time: 40.4 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "random_forest_classifier_prediction(label__X_valid, y_valid)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "19baa266",
   "metadata": {},
   "source": [
    "### Catboost Classifier with hyperparameter tuning \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "id": "574a4a3f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The best hyperparameters are: {'depth': 4, 'l2_leaf_reg': 7, 'learning_rate': 0.5}\n",
      "CPU times: user 8min 30s, sys: 4.54 s, total: 8min 34s\n",
      "Wall time: 8min 54s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "grid_cb = {'learning_rate': [0.001, 0.01, 0.5],\n",
    "        'depth': [4, 6, 10],\n",
    "        'l2_leaf_reg': [1, 3, 5, 7, 9]\n",
    "       }\n",
    "cb_clf = CatBoostClassifier(\n",
    "    iterations=200,\n",
    "    cat_features = object_cols,\n",
    "    logging_level = 'Silent',\n",
    "    eval_metric='AUC',\n",
    "    early_stopping_rounds = 50,\n",
    "    random_state = 12345)\n",
    "\n",
    "grid_search_cb = GridSearchCV(estimator = cb_clf, param_grid = grid_cb, scoring=\"roc_auc\", cv=5)\n",
    "\n",
    "cb_cv = grid_search_cb.fit(X_train, y_train)\n",
    "\n",
    "print('The best hyperparameters are: {}'.format(cb_cv.best_params_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "id": "be06e001",
   "metadata": {},
   "outputs": [],
   "source": [
    "def train_catboost_classifier(X_train, y_train, X_valid, y_valid):\n",
    "    \"\"\"This function trains a catboost classifier model\"\"\"\n",
    "    global cb_model\n",
    "    \n",
    "    cb_model = CatBoostClassifier(**grid_search_cb.best_params_)\n",
    "    cb_model.fit(X_train, y_train, cat_features = object_cols, eval_set=(X_valid, y_valid), verbose=False, plot=False) # train the model \n",
    "    \n",
    "def catboost_classifier_prediction(X_valid, y_valid):\n",
    "    \"\"\"\n",
    "    This function is used to make prediction \n",
    "    using the catboost classifier model\n",
    "    \"\"\"\n",
    "    global cb_pred, cb_pred_proba\n",
    "    cb_pred = cb_model.predict(X_valid)\n",
    "    cb_pred_proba = cb_model.predict_proba(X_valid)[:, 1]\n",
    "   \n",
    "    print(\"\\033[1m\" + 'AUC-ROC Score and Accuracy using CatBoost Classifier' + \"\\033[0m\")\n",
    "    print('AUC-ROC Score: {:.2%}'.format(roc_auc_score(y_valid, cb_pred_proba)))\n",
    "   \n",
    "    cb_accuracy = accuracy_score(y_valid, cb_pred)\n",
    "    print('Accuracy score: {:.2%}'.format(cb_accuracy))\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "id": "cd2039bb",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 5.26 s, sys: 63.9 ms, total: 5.33 s\n",
      "Wall time: 5.49 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "train_catboost_classifier(X_train, y_train, X_valid, y_valid)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "id": "349df12e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1mAUC-ROC Score and Accuracy using CatBoost Classifier\u001b[0m\n",
      "AUC-ROC Score: 91.55%\n",
      "Accuracy score: 86.68%\n",
      "CPU times: user 21.7 ms, sys: 31 µs, total: 21.7 ms\n",
      "Wall time: 30.9 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "catboost_classifier_prediction(X_valid, y_valid)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8ae95d22",
   "metadata": {},
   "source": [
    "AUC-ROC and accuracy are looking much better with this boosted model. Lets try one more."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "640a743e",
   "metadata": {},
   "source": [
    "### LightGBM Classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "id": "eb352c52",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The best hyperparameters are: {'learning_rate': 0.05, 'n_estimators': 500, 'num_leaves': 5}\n",
      "CPU times: user 26 s, sys: 259 ms, total: 26.3 s\n",
      "Wall time: 26.5 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "grid_lgbm = {'learning_rate': [0.05, 0.1],\n",
    "             'n_estimators': [50, 500],\n",
    "             'num_leaves': [5, 10]\n",
    "            }\n",
    "\n",
    "lgbm_clf = LGBMClassifier(random_state = 12345)\n",
    "\n",
    "grid_search_lgbm = GridSearchCV(\n",
    "    estimator = lgbm_clf, \n",
    "    param_grid = grid_lgbm, \n",
    "    scoring = \"accuracy\", \n",
    "    cv = 5, \n",
    "    n_jobs = 1\n",
    ")\n",
    "\n",
    "lgbm_cv = grid_search_lgbm.fit(label__X_train, y_train)\n",
    "\n",
    "print('The best hyperparameters are: {}'.format(lgbm_cv.best_params_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "id": "d941797e",
   "metadata": {},
   "outputs": [],
   "source": [
    "def train_lightGBM_classifier(X_train, y_train):\n",
    "    \"\"\"This function trains a LightGBM classifier\"\"\"\n",
    "    global lgbm_model\n",
    "    lgbm_model = LGBMClassifier(**grid_search_lgbm.best_params_)\n",
    "    lgbm_model.fit(X_train, y_train) \n",
    "    \n",
    "def lightGBM_classifier_prediction(X_valid, y_valid):\n",
    "    \"\"\"\n",
    "    This function is used to make prediction \n",
    "    using the LightGBM classifier model\n",
    "    \"\"\"\n",
    "    global lgbm_pred, lgbm_pred_proba\n",
    "    lgbm_pred = lgbm_model.predict(X_valid)\n",
    "    lgbm_pred_proba = lgbm_model.predict_proba(X_valid)[:, 1]\n",
    "    \n",
    "    print(\"\\033[1m\" + 'AUC-ROC Score and Accuracy using LightGBM Classifier' + \"\\033[0m\")\n",
    "    print('AUC-ROC Score: {:.2%}'.format(roc_auc_score(y_valid, lgbm_pred_proba)))\n",
    "\n",
    "    lgbm_accuracy = accuracy_score(y_valid, lgbm_pred)\n",
    "    print('Accuracy score: {:.2%}'.format(lgbm_accuracy))\n",
    " "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f54e4a8b",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\">\n",
    "<b>Reviewer's comment</b>\n",
    "\n",
    "In `*_classifier_prediction()` functions ROC-AUC is calculated correctly\n",
    "\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "id": "a3b79fa1",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 594 ms, sys: 7.03 ms, total: 601 ms\n",
      "Wall time: 590 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "train_lightGBM_classifier(label__X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "id": "362ee505",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1mAUC-ROC Score and Accuracy using LightGBM Classifier\u001b[0m\n",
      "AUC-ROC Score: 91.04%\n",
      "Accuracy score: 87.66%\n",
      "CPU times: user 82 ms, sys: 3.92 ms, total: 85.9 ms\n",
      "Wall time: 26.5 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "lightGBM_classifier_prediction(label__X_valid, y_valid)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ff18eee0",
   "metadata": {},
   "source": [
    "We can see that lightGBM achieved a AUC-ROC score of 91% and an accuracy score of 87%. Lets move on to chosing our algorithm.  "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f4279f0c",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\">\n",
    "<b>Reviewer's comment</b>\n",
    "\n",
    "Great, you tried a few different models and tuned their hyperparameters using cross-validated grid search\n",
    "\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c8b5826e",
   "metadata": {},
   "source": [
    "### Final testing "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "id": "14447668",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best model is CatBoost with a score of 0.9053514318211308\n",
      "Best params is : {'depth': 4, 'l2_leaf_reg': 7, 'learning_rate': 0.5}\n"
     ]
    }
   ],
   "source": [
    "# determine best algorithm\n",
    "models = {'LogisticRegression':logreg_cv.best_score_,\n",
    "          'RandomForest':rf_cv.best_score_,\n",
    "          'LightGBM': lgbm_cv.best_score_,\n",
    "          'CatBoost': cb_cv.best_score_\n",
    "         }\n",
    "\n",
    "bestalgorithm = max(models, key=models.get)\n",
    "print('Best model is', bestalgorithm,'with a score of', models[bestalgorithm])\n",
    "if bestalgorithm == 'LogisticRegression':\n",
    "    print('Best params is :', logreg_cv.best_params_)\n",
    "if bestalgorithm == 'RandomForest':\n",
    "    print('Best params is :', rf_cv.best_params_)\n",
    "if bestalgorithm == 'LightGBM':\n",
    "    print('Best params is :', lgbm_cv.best_params_)\n",
    "if bestalgorithm == 'CatBoost':\n",
    "    print('Best params is :', cb_cv.best_params_)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "99b0a04d",
   "metadata": {},
   "source": [
    "### Predictions on the test dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "id": "12c6c94c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1mAUC-ROC Score and Accuracy using CatBoost Classifier\u001b[0m\n",
      "AUC-ROC Score: 90.92%\n",
      "Accuracy score: 86.83%\n",
      "CPU times: user 22.8 ms, sys: 45 µs, total: 22.8 ms\n",
      "Wall time: 19.8 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "#final test with the test dataset\n",
    "catboost_classifier_prediction(X_test, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "id": "3db16d84",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1mF1 score: \u001b[0m 0.734\n",
      "\u001b[1mAccuracy Score: \u001b[0m 86.83%\n",
      "\u001b[1mPrecision: \u001b[0m 0.804\n",
      "\u001b[1mRecall: \u001b[0m 0.675\n",
      "\n",
      "\u001b[1mConfusion Matrix\u001b[0m\n",
      "--------------------------------------------------\n",
      "[[1209   78]\n",
      " [ 154  320]]\n",
      "\n",
      "\u001b[1mClassification report\u001b[0m\n",
      "--------------------------------------------------\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.89      0.94      0.91      1287\n",
      "           1       0.80      0.68      0.73       474\n",
      "\n",
      "    accuracy                           0.87      1761\n",
      "   macro avg       0.85      0.81      0.82      1761\n",
      "weighted avg       0.86      0.87      0.86      1761\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# function to calculate model evaluation metrics\n",
    "def final_model_evaluation(y_test, test_predictions):\n",
    "    print(\"\\033[1m\" + 'F1 score: ' + \"\\033[0m\", '{:.3f}'.format(f1_score(y_test, test_predictions)))\n",
    "    print(\"\\033[1m\" + 'Accuracy Score: ' + \"\\033[0m\", '{:.2%}'.format(accuracy_score(y_test, test_predictions)))\n",
    "    print(\"\\033[1m\" + 'Precision: ' + \"\\033[0m\", '{:.3f}'.format(precision_score(y_test, test_predictions)))\n",
    "    print(\"\\033[1m\" + 'Recall: ' + \"\\033[0m\", '{:.3f}'.format(recall_score(y_test, test_predictions)))\n",
    "    print()\n",
    "    print(\"\\033[1m\" + 'Confusion Matrix' + \"\\033[0m\")\n",
    "    print('-'*50)\n",
    "    print(confusion_matrix(y_test, test_predictions))\n",
    "    print()\n",
    "    print(\"\\033[1m\" + 'Classification report' + \"\\033[0m\")\n",
    "    print('-'*50)\n",
    "    print(classification_report(y_test, test_predictions))\n",
    "    print()\n",
    "\n",
    "final_model_evaluation(y_test, cb_pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ae1f4639",
   "metadata": {},
   "source": [
    "<div class=\"alert alert-success\">\n",
    "<b>Reviewer's comment</b>\n",
    "\n",
    "The final model was evaluated on the test set for an unbiased estimate of its generalization performance\n",
    "    \n",
    "\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ed113598",
   "metadata": {},
   "source": [
    "## Conclusions"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "eedd597d",
   "metadata": {},
   "source": [
    "Through exploratry data analysis we were able to learn useful information about customer churn. These insights were: \n",
    "- Most customers prefer the month-to- month plan, with a majority of 61% of the customer base chosing this option\n",
    "- Customers with a 2 year contract are bringing in more revenue than the other two contract options \n",
    "- Customers on a 2 year contract are churning less than the other contracts\n",
    "- customers having 5 or more services with telecom are less likely to churn\n",
    "Then through machine learning we were able to develop a model which is able to predict customer churn with 87% accuracy and a AUR-ROC score of 90.9. This will be useful to telecom in reducing customer churn by having a tool to predict it.\n",
    "\n",
    "Action steps for Telecom would be the following:\n",
    "- incentivize 2 year contract participation, one month free, get two services for the price of one for the first year etc.\n",
    "- boost multiple service use by using promotions \n",
    "- introduce bonuses and incentives before peak churn months\n",
    "- utilize model to predict customers likelyhood of canceling and use that information to target incentives, reducing the number of promotions distributed and therefore saving the company money\n",
    " "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "cf47808d",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.5"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": true,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": true
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}