Introduction

Welcome to simple classification project usin KNN algorithm | K - Nearest Neighbor algorithm

• Fristly, let thank for contributing the "Bad words" data of many highschool students at Danang city.
• Secondly, I hope you will have the funny time in this project. I had tried editing and customizing my code in the easiest way, which can help you understand KNN algorithm and how to implement it on a project.

In summary, it also contained Auto Find the best K algorithm for KNN. A powerful tool that can save your time. The best K is K:

• The closest neighbor can affect the test accuracy on all your data the lowest
• It is NOT 1 - because you need to be objective
• It is NOT even number - because you use the voting in KNN, the label of the new point is what the label has more quantity, so if you have 2 blue and 2 red, the system will be confusing.
• A second largest number

A few notes about classification_Vietnamese_sentence_knn_simple project

• It's example just fun. But you can use it in spam or impolite classification.

• You should run find_k.js to get best k before run testtrainwithknn.js.

• Only use javascript and don't contain any library to build it.

• I hope it can help you fun and learn KNN easier.

• Try it, Good luck !!!!

The flow

If you want to try with your data or do not understand how it work, please see it in this picture, you can see its flow.

You need preparing data first (bad words list, types of sentences) and find out the best K. Finnal, use the best K has been found for testing. The main flow is the blue lines.

Data classified sentences ready for KNN

Sentence	Label
Thứ phò cẩu	Câu bị nghi vấn
Con cẩu sanh	Câu bị nghi vấn
Bươm bướm dối trá không thấy ngượng mồm	Câu bị nghi vấn
Nhớ gia đình ạ	Câu bình thường
Bản thân của ngày xưa	Câu bình thường
Gặt hái được nhiều thành công	Câu bình thường
May mắn lắm đó nha	Câu bình thường
Tui nói ông nghe nè, tui yêu ông	Câu bình thường
Thương em là điều anh không thể ngờ	Câu bình thường
em muốn uống trà sữa	Câu bình thường
xịt sữa	Câu bị nghi vấn
bình thường, cx ko hay chi mấy	Câu bình thường
Thay ava chào năm mới cho đỡ mốc	Câu bình thường
..See more at datasentence.js...	...See more at datasentence.js..

The structure of the sentence point

You need to convert sentences into points on a spatial graph (2D - (x, y)) before using KNN to classify. So, how I did it? Follow this caculation formula:

x = len(bad words in sentence)/len(sentence)

y = len(bad words in sentence)

In the x, I removed duplicate bad words in each sentence before dividing.

And I used the Euclidean distance to calculate the relationship between two points.

Bad words

Word
cc
đồ chó
clon
đm
địt con mẹ
địt mẹ
chu cái mỏ lồn
súc vật
Đạp mày nát bét
Đéo
Đè bẹp mày
Khinh tao à
...See more at arraytrainbot.js...

List K test - get it in file find_k.js

K	Correct prediction
1	74
2	69
3	74
4	72
5	74
6	71
7	71
8	71
9	71
10	71

That is result for find_k.js : best K : 5 . It isn't good but not bad. That was an objective result in 10 test rounds
The accuracy is 
- 74% for 2237 bad words* @@ 
*: (word + sentence contain bad words) and 85 classified sentences