700字范文,内容丰富有趣,生活中的好帮手!
700字范文 > 【Machine Learning in Action --4】朴素贝叶斯电子邮件垃圾过滤

【Machine Learning in Action --4】朴素贝叶斯电子邮件垃圾过滤

时间:2020-02-06 15:24:00

相关推荐

【Machine Learning in Action --4】朴素贝叶斯电子邮件垃圾过滤

摘要:这里用的是词袋模型,即一个词在文档中出现不止一次,每个单词可以出现多次。

1、准备数据:切分文本

前一节过滤网站恶意留言中词向量是给定的,下面介绍如何从文本文档中构建自己的词列表

先举例说明,在python提示符下输入:

>>> mySent='This book is the best book on python or M.L. I have ever laid eyes upon.'>>> mySent.split()['This', 'book', 'is', 'the', 'best', 'book', 'on', 'python', 'or', 'M.L.', 'I', 'have', 'ever', 'laid', 'eyes', 'upon.']#标点符号也被当成了词的一部分,可以使用正则表达式来切分句子,其中分隔符是除单词、数字外的任意字符串>>> import re>>> regEx=pile('\\W*')>>> listOfTokens=regEx.split(mySent)>>> listOfTokens['This', 'book', 'is', 'the', 'best', 'book', 'on', 'python', 'or', 'M', 'L', 'I', 'have', 'ever', 'laid', 'eyes', 'upon', '']#去掉空字符串,通过计算每个字符串的长度,只返回长度大于0的字符串>>> [tok for tok in listOfTokens if len(tok)>0]['This', 'book', 'is', 'the', 'best', 'book', 'on', 'python', 'or', 'M', 'L', 'I', 'have', 'ever', 'laid', 'eyes', 'upon']#将字符串全部转换成小写(.lower())或者大写(.upper())>>> [tok.lower() for tok in listOfTokens if len(tok)>0]['this', 'book', 'is', 'the', 'best', 'book', 'on', 'python', 'or', 'm', 'l', 'i', 'have', 'ever', 'laid', 'eyes', 'upon']

本例中共有50封电子邮件,采用的是email文件夹下的ham文件和spam文件,其中ham文件下有25份d.txt(d是1到25)文件,spam文件下也有25份d.txt(d是1到25)文件。其中的10封邮件被随机选择为测试集。以下分别是ham文件下1.txt的内容,spam文件下1.txt的内容:

创建一个bayes.py文件,添加以下代码:

#!/usr/bin/python

#-*- coding:utf-8 -*-

#from numpy import *

#创建一个包含在所有文档中出现的不重复词的列表def createVocabList(dataSet):vocabSet=set([]) #创建一个空集for document in dataSet:vocabSet=vocabSet|set(document) #创建两个集合的并集return list(vocabSet)#该函数输入参数为词汇表及其某个文档,输出是文档向量def setOfWords2Vec(vocabList,inputSet):returnVec=[0]*len(vocabList)for word in inputSet:if word in inputSet:returnVec[vocabList.index(word)]+=1 #这里是词袋模型,与词集模型不一样else:print "the word:%s is not in my Vocabulary!" % wordreturn returnVec

2、训练算法:从词向量计算概率

前面介绍了如何将一组单词转换为一组数字,接下来看看如何使用这些数字计算概率。现在已经知道一个词是否出现在一篇文档中,也知道该文档所属类别。

(1)

首先通过类别i中文档数除以总的文档数来计算概率P(Ci),然后计算P(w|Ci),即P(w0|Ci)P(w1Ci)...P(wN|Ci)来计算上述概率。

#朴素贝叶斯分类器训练函数def trainNBO(trainMatrix,trainCategory):numTrainDocs=len(trainMatrix)numWords=len(trainMatrix[0])pAbusive=sum(trainCategory)/float(numTrainDocs)p0Num=ones(numWords);p1Num=ones(numWords) #计算p(w0|1)p(w1|1),避免其中一个概率值为0,最后的乘积为0p0Demo=2.0;p1Demo=2.0 #初始化概率for i in range(numTrainDocs):if trainCategory[i]==1:p1Num+=trainMatrix[i]p1Demo+=sum(trainMatrix[i])else:p0Num+=trainMatrix[i]p0Demo+=sum(trainMatrix[i])#p1Vect=p1Num/p1Demo#p0Vect=p0Num/p0Demop1Vect=log(p1Num/p1Demo) #计算p(w0|1)p(w1|1)时,大部分因子都非常小,程序会下溢出或得不到正确答案(相乘许多很小数,最后四舍五入会得到0)p0Vect=log(p0Num/p0Demo)return p0Vect,p1Vect,pAbusive

3、测试算法:使用朴素贝叶斯进行交叉验证

#朴素贝叶斯分类函数def classifyNB(vec2Classify,p0Vec,p1Vec,pClass1):p1=sum(vec2Classify*p1Vec)+log(pClass1)p0=sum(vec2Classify*p0Vec)+log(1.0-pClass1)if p1>p0:return 1else:return 0#文件解析及完整的垃圾邮件测试函数def textParse(bigString):import relistOfTokens=re.split(r'\W*',bigString)return [tok.lower() for tok in listOfTokens if len(tok)>2]def spamTest():docList=[];classList=[];fullText=[]for i in range(1,26):wordList=textParse(open('email/spam/%d.txt'% i).read())docList.append(wordList)fullText.extend(wordList)classList.append(1)wordList=textParse(open('email/ham/%d.txt'% i).read())docList.append(wordList)fullText.extend(wordList)classList.append(0)vocabList=createVocabList(docList)trainingSet=range(50);testSet=[] #trainingSet是一个整数列表,其中的值从0到49for i in range(10): #随机选择其中10个文件randIndex=int(random.uniform(0,len(trainingSet)))testSet.append(trainingSet[randIndex]) #选择出的数字所对应的文档被添加到测试集del(trainingSet[randIndex]) #同时被选中的数据将从训练集中踢除trainMat=[];trainClasses=[]for docIndex in trainingSet:trainMat.append(setOfWords2Vec(vocabList,docList[docIndex]))trainClasses.append(classList[docIndex])p0V,p1V,pSpam=trainNBO(array(trainMat),array(trainClasses))errorCount=0for docIndex in testSet:wordVector=setOfWords2Vec(vocabList,docList[docIndex])if classifyNB(array(wordVector),p0V,p1V,pSpam)!=classList[docIndex]:errorCount+=1print 'the error rate is:',float(errorCount)/len(testSet)

下面对上述过程进行尝试,在python提示符下输入:

>>> reload(bayes)<module 'bayes' from 'bayes.py'>>>> bayes.spamTest()the error rate is: 0.1>>> bayes.spamTest()the error rate is: 0.0

函数spamTest()会输出在10封随机选择的电子邮件上的分类错误绿。既然这些电子邮件是随机选择的,所以每次的输出结果可能有些差别。

解释:

>>> docList=[];classList=[];fullText=[]>>> for i in range(1,26):...wordList=bayes.textParse(open('email/spam/%d.txt'%i).read())...docList.append(wordList)...fullText.extend(wordList)...classList.append(1)...>>> wordList['experience', 'with', 'biggerpenis', 'today', 'grow', 'inches', 'more', 'the', 'safest', 'most', 'effective', 'methods', 'of_penisen1argement', 'save', 'your', 'time', 'and', 'money', 'bettererections', 'with', 'effective', 'ma1eenhancement', 'products', 'ma1eenhancement', 'supplement', 'trusted', 'millions', 'buy', 'today']>>> docList[['codeine', '15mg', 'for',..., 'buy', 'today']]>>> fullText['codeine', '15mg', 'for', ...,'buy', 'today']>>> classList[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]>>> len(wordList)29>>> len(docList)25>>> len(fullText)795>>> len(classList)25

>>> docList=[];classList=[];fullText=[]>>> for i in range(1,26):...wordList=bayes.textParse(open('email/ham/%d.txt'%i).read())...docList.append(wordList)...fullText.extend(wordList)...classList.append(0)... >>> wordList['that', 'cold', 'there', 'going', 'retirement', 'party', 'are', 'the', 'leaves', 'changing', 'color']>>> docList[['codeine', '15mg', ..., 'changing', 'color']]>>> fullText['codeine', '15mg', ..., 'changing', 'color']>>> classList[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]>>> len(wordList)11>>> len(docList)25>>> len(fullText)967>>> len(classList)25

>>> docList=[];classList=[];fullText=[]>>> for i in range(1,26):...wordList=bayes.textParse(open('email/spam/%d.txt'%i).read())...docList.append(wordList)...fullText.extend(wordList)...classList.append(1)...wordList=bayes.textParse(open('email/ham/%d.txt'%i).read())...docList.append(wordList)...fullText.extend(wordList)...classList.append(0)... >>> wordList['that', 'cold', 'there', 'going', 'retirement', 'party', 'are', 'the', 'leaves', 'changing', 'color']>>> docList[['codeine', '15mg', ..., 'changing', 'color']]>>> fullText['codeine', '15mg', ..., 'changing', 'color']>>> classList[1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0]>>> len(wordList)11>>> len(docList)50>>> len(fullText)1762>>> len(classList)50

>>> vocabList=bayes.createVocabList(docList)>>> vocabList['all', 'code', ..., 'others', 'once']>>> len(vocabList)692>>> trainingSet=range(50);testSet=[]>>> for i in range(10):...randIndex=int(random.uniform(0,len(trainingSet)))...testSet.append(trainingSet[randIndex])...del(trainingSet[randIndex])... >>> trainMat=[];trainClasses=[]>>> for docIndex in trainingSet:...trainMat.append(bayes.setOfWords2Vec(vocabList,docList[docIndex]))...trainClasses.append(classList[docIndex])... >>> shape(trainMat)(40, 692)#表示40行692列,即40篇训练文档,692个不重复的词汇

>>> p0V,p1V,pSpam=bayes.trainNBO(array(trainMat),array(trainClasses))>>> len(p0V)692>>> len(p1V)692

这里一直出现的错误是将垃圾邮件误判为正常邮件,相比之下,将垃圾邮件误判为正常邮件要比正常邮件归到垃圾邮件好。

本内容不代表本网观点和政治立场,如有侵犯你的权益请联系我们处理。
网友评论
网友评论仅供其表达个人看法,并不表明网站立场。