机器学习实战专题（四）Naive Bayes

 使用朴素贝叶斯过滤垃圾邮件

---

   说明:  将 `email` 文件夹放在当前目录下。

  1.词集模型

 

## 1 词集模型

import numpy as np
import pandas as pd
import re


# 1.1文件解析及完整的垃圾邮件测试




def textParse(text):
    listofTokens=re.split(r'\w+',text)
    return [tok.lower() for tok in listofTokens if len(tok)>2]





def createWordList(dataSet):
    wordSet=set([])
    for dataList in dataSet:
        wordSet=wordSet|set(dataList)
    return list(wordSet)





def setOfWords2Vec(wordList,inputSet):
    returnVec=[0]*len(wordList)
    for word in inputSet:
        if word in wordList:
            returnVec[wordList.index(word)]=1
        else:
            print('the word:%s is not in my Vocabulary!'%word)
    return returnVec


# 1.2 朴素贝叶斯模型训练




def trainNB(trainMatrix,trainCategory):
    numTrainDocs=len(trainMatrix)
    numWords=len(trainMatrix[0])
    pClass1=sum(trainCategory)/float(numTrainDocs)
    p0Num=np.ones(numWords)
    p1Num=np.ones(numWords)
    p0sum=2.0
    p1sum=2.0
    
    for index in range(numTrainDocs):
        if trainCategory[inedx]==1:
            p1Num+=trainMatrix[index]
            p1Sum+=1
        else:
            p0Num+=trainMatrix[index]
            p0Sum+=1
    p0Vect=np.log(p0Num/p0Sum)
    p1Vect=np.log(p1Num/p1Sum)
    
    p0AntiVect=np.log(1-p0Num/p0Sum)
    p1AntiVect=np.log(1-p0Num/p0Sum)
    
    return pClass1,p0Vect,p1Vect,p0AntiVect,p1AntiVect
            





def classifyNB(classifyData,pClass1,p0Vec,p1Vec,p0Antiv,p1Antiv):
    antiClassifyData=1-classifyData
    p1=np.log(pClass1)+sum(classifyData*p1Vec)+sum(antiClassifyData*p1AntiV)
    p0=np.log(1.0-pClass1)+sum(classifyData*p0Vec)+sum(antiClassifyData*p0AntiV)
    if p1>p0:
        return 1
    else:
        return 0





def testSetOfWordsModel():
    docList=[]
    classList=[]
    fullText=[]
    
    for index in range(1,26):
        wordList=textParse(open('email/spam/%d.txt'% index).read())
        docList.append(wordList)
        fullText.extend(wordList)
        classList.append(1)
        wordList=textParse(open('email/ham/%d.txt'% index).read())
        docList.append(wordList)
        fullText.extend(wordList)
        classList.append(0)
    myWordList=createWordList(docList)
    
    trainingSet=list(range(50))
    testSet=[]
    
    for index in range(10):
        randIndex=int(np.random.uniform(0,len(trainingSet)))
        testSet.append(trainingSet[randIndex])
        del(trainingSet[randIndex])
    
    trainMatrix=[]
    trainClass=[]
    
    for docIndex in trainingSet:
        trainMatrix.append(setOfWords2Vec(myWordList,docList[docIndex]))
        trainClass.append(classList[docIndex])
        
    pClass1,p0V,p1V,p0AntiV,p1AntiV=trainNB(trainMatrix,trainClass)
    
    errorcount=0
    for docIndex in testSet:
        wordVector=np.array(setOfWords2Vec(myWordList,docList[docIndex]))
        if classifyNB(wordVector,pClass1,p0V,p1V,p0AntiV,p1AntiV)!=classList[docIndex]:
            errorcount+=1
    

 2 词袋模型

# #  2.词袋模型




import numpy as np
import pandas as pd
import re





def textParse(text):
    listofTokens=re.split(r'\w+',text)
    return [tok.lower() for tok in listofTokens if len(tok)>2]





def createWordList(dataSet):
    wordSet=set([])
    for dataList in dataSet:
        wordSet=wordSet|set(dataList)
    return list(wordSet)





def bagOfWords2Vec(wordList,inputSet):
    returnVec=[0]*len(wordList)
    for word in inputSet:
        if word in inputSet:
            if word in wordList:
                returnVec[wordList.index(word)]+=1
    return returnVec





def trainNB(trainMatrix,trainCategory):
    numTrainDocs=len(trainMatrix)
    numWords=len(trainMatrix[0])
    p0Num=np.ones(numWords)
    p1Num=np.ones(numWords)
    p0Sum=0.0
    p1Sum=0.0
    
    for index in range(numTrainDocs):
        if trainCategory[index]==1:
            p1Num+=trainMatrix[index]
            p1Sum+=sum(trainMatrix[index])
        else:
            p0Num+=trainMatrix[index]
            p0Sum+=sum(trainMatrix[index])
        
    pClass1=p1Sum/float(p1Sum + p0Sum)
    p0Vect=np.log(p0Num/(p0Sum+numWords))
    p1Vect=np.log(p1Num/(p1Sum+numWords))
    
    return pClass1,p0Vect,p1Vect





def classifyNB(classifyData,pClass1,p0Vec,p1Vec):
    p1=sum(classifyData*p1Vec)+np.log(pClass1)
    p0=sum(classifyData*p0Vec)+np.log(1.0-pClass1)
    if p1>p0:
        return 1
    else:
        return 0





def testBagOfWordsModel():
    docList=[]
    classList=[]
    fullText=[]
    
    for index in range(1,26):
        wordList=textParse(open('email/spam/%d.txt'% index).read())
        docList.append(wordList)
        fullText.extend(wordList)
        classList.append(1)
        wordList=textParse(open('email/ham/%d.txt'% index).read())
        docList.append(wordList)
        fullText.extend(wordList)
        classList.append(0)
    myWordList=createWordList(docList)
    
    trainingSet=list(range(50))
    testSet=[]
    
    for index in range(10):
        randIndex=int(np.random.uniform(0,len(trainingSet)))
        testSet.append(trainingSet[randIndex])
        del(trainingSet[randIndex])
    
    trainMatrix=[]
    trainClass=[]
    
    for docIndex in trainingSet:
        trainMatrix.append(bagOfWords2Vec(myWordList,docList[docIndex]))
        trainClass.append(classList[docIndex])
        
    pClass1,p0V,p1V=trainNB(trainMatrix,trainClass)
    
    errorcount=0
    for docIndex in testSet:
        wordVector=np.array(bagOfWords2Vec(myWordList,docList[docIndex]))
        if classifyNB(wordVector,pClass1,p0V,p1V)!=classList[docIndex]:
            errorcount+=1