ASR.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "**Implementation of Approximate Stream Regression**\n",
    "\n",
    "\n",
    "While recieving stream data we need to continously update the value of parameters as we recieve it. $X_k$ is the data that is recieved in the $k^{th}$ window.\n",
    "\n",
    "\n",
    "$\\beta_k ^{'} = (X_k^TX_k)^{-1}X^T_ky_k$\n",
    "\n",
    "$\\beta_k = (1-\\alpha)\\beta_k ^{'}+ (\\alpha)\\beta_{k-1}$\n",
    "\n",
    "Here $\\alpha$ is Smoothing factor \\\\\n",
    "$\\beta_k^{'}$ is Parameter vector for the kth window calculated considering only the data records of kth window. \\\\\n",
    "\n",
    "$\\beta_{k-1}$ is Parameter vector for the kth window calculated considering all data records seen upto (k-1)th window. \\\\\n",
    "\n",
    "$\\beta_{k}$ is Parameter vector for the kth window calculated considering all data records seen upto (k)th window. \\\\\n",
    "\n",
    "**Importing Libaries**\n",
    "\n",
    "from sklearn import metrics\n",
    "import pandas as pd\n",
    "\n",
    "**Loading the dataset and splitting the test set**\n",
    "\n",
    "count=0\n",
    "\n",
    "testdataframe=pd.read_csv(\"boston.csv\",header=None)\n",
    "testdataarray=testdataframe.values\n",
    "X_test=testdataarray[400:506,:-1] \n",
    "y_test=testdataarray[400:506,-1]\n",
    "#print(y_test)\n",
    "\n",
    "**Simulation of Streaming Data**\n",
    "\n",
    "**Function for generating stream data**\n",
    "\n",
    "\n",
    "**Description : This function generates batches of data with size equal to args:(wsize) and instead of using return uses yield which will**\n",
    "\n",
    "**enable to generate next batch of training data each time it is iterated**\n",
    "\n",
    "args : wsize\n",
    "\n",
    "yield : stream data\n",
    "\n",
    "def myGeneratorfun(wsize):\n",
    "    print(\"In myGeneratorfun:\")\n",
    "    counter=0\n",
    "    for chunk in pd.read_csv('boston.csv', header=None, chunksize=wsize):\n",
    "        print(\"In for loop, counter vlaue is\", counter)\n",
    "        chunk_array=chunk.values\n",
    "        counter=counter+1\n",
    "        yield (chunk_array[:,:-1], chunk_array[:,-1])\n",
    "        if counter >= 8:\n",
    "            break\n",
    "\n",
    "**Calling stream data genrator with wsize = 50**\n",
    "\n",
    "myGenerator = myGeneratorfun(50)\n",
    "\n",
    "print(X_test.shape)\n",
    "\n",
    "**Steps for ASR**\n",
    "\n",
    "\n",
    "M=np.zeros((X_test.shape[1],X_test.shape[1]))\n",
    "\n",
    "V=np.zeros((X_test.shape[1]))\n",
    "Beta=np.zeros(X_test.shape[1])\n",
    "\n",
    "alpha = 0.75\n",
    "Beta_k = Beta\n",
    "\n",
    "flag = 0\n",
    "\n",
    "for (x,y) in myGenerator:\n",
    "\n",
    "    try:\n",
    "        XTXi=inv(np.dot(x.T,x))\n",
    "    except LinAlgError:\n",
    "        XTXi=pinv(np.dot(x.T,x))\n",
    "\n",
    "    XTy=np.dot(x.T,y)\n",
    "\n",
    "    Beta_k_dash=np.dot(XTXi,XTy)\n",
    "\n",
    "    if flag==0 :\n",
    "        Beta_k = Beta_k_dash\n",
    "        flag=1\n",
    "    else :\n",
    "        Beta_k = (1-alpha)*Beta_k_dash + alpha*Beta_k\n",
    "\n",
    "    print(\"Beta':\", Beta_k_dash)\n",
    "    print(\"Betas:\", Beta_k)\n",
    "    print(\"Betas Shape:\", Beta.shape)   \n",
    "\n",
    "    predictions=np.dot(Beta_k,X_test.T)\n",
    "    #print(predictions.shape)\n",
    "    print(\"MAE:\", metrics.mean_absolute_error(y_true=y_test,y_pred=predictions))\n",
    "    print(\"MSE:\", metrics.mean_squared_error(y_true=y_test,y_pred=predictions))\n",
    "\n",
    "M.shape"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}