9. naive Bayes

Introduction to naive Bayes （ Probability Classification ）

Probability basis

joint probability 、 Conditional probability and mutual independence

Bayes' formula

Introduce

Case study

API

Emotional analysis of commodity reviews

Import dependence

import pandas as pd
import numpy as py
import jieba
import matplotlib.pyplot as plt
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.naive_bayes import MultinomialNB

get data

data = pd.read_csv("evaluation.csv",encoding="gbk")
data

Basic data processing

#  Take out the content column , For later analysis 
content = data[" Content "]
content

#  Convert the positive and negative comments in the evaluation into figures 
data.loc[data.loc[:," evaluation "] == " Praise "," Comment number "] =1
data.loc[data.loc[:," evaluation "] == " Bad review "," Comment number "] =0
data

#  Choose a stop word 
stopwords = []
with open("stopwords.txt","r",encoding="utf-8") as f:
    lines = f.readlines()
    for tmp in lines:
        line = tmp.strip()
        stopwords.append(line)
stopwords = list(set(stopwords))
print(" Stop words :\n",stopwords)

#  Convert content to standard format 
comment_list = []
for tmp in content:
    #  Cut words into words 
    seg_list = jieba.cut(tmp,cut_all=False)
    seg_str = ",".join(seg_list)
    comment_list.append(seg_str)
comment_list

#  Count the number of words 
con = CountVectorizer(stop_words=stopwords)
X = con.fit_transform(comment_list)
X.toarray()

##  Prepare training and test sets 
x_train = X.toarray()[:10,:]
y_train = data[" evaluation "][:10]
print(" Training set :\n",x_train)
print(" Training set :\n",y_train)

x_test = X.toarray()[10:,:]
y_test = data[" evaluation "][10:]
print(" Test set :\n",x_test)
print(" Test set :\n",y_test)

model training

mb = MultinomialNB(alpha=1)
mb.fit(x_train,y_train)
y_pre = mb.predict(x_test)

Model to evaluate

print(" Predictive value ",y_pre)
print(" True value ",y_test)

mb.score(x_test,y_test)

Advantages and disadvantages of naive Bayes

Naive Bayesian content summary

NB Principle

Naive Bayes where is simplicity

Why introduce the conditional independence assumption

In estimating conditional probability P(X|Y) The probability of occurrence is 0 How to deal with the situation of

Why is the assumption of attribute independence difficult to hold in practice , But naive Bayes can still achieve better results

Naive Bayes and LR（ Logical regression ） The difference between