朴素贝叶斯算法优缺点优点：在数据较少的情况下依然有效，可以处理多类别问题缺点：对输入数据的准备方式敏感适用数据类型：标称型数据算法思想：朴素贝叶斯
比如我们想判断一个邮件是不是垃圾邮件，那么我们知道的是这个邮件中的词的分布，那么我们还要知道：垃圾邮件中某些词的出现是多少，就可以利用贝叶斯定理得到。
朴素贝叶斯分类器中的一个假设是：每个特征同等重要
函数loadDataSet()
创建数据集，这里的数据集是已经拆分好的单词组成的句子，表示的是某论坛的用户评论，标签1表示这个是骂人的createVocabList(dataSet)
找出这些句子中总共有多少单词，以确定我们词向量的大小setOfWords2Vec(vocabList, inputSet)
将句子根据其中的单词转成向量，这里用的是伯努利模型，即只考虑这个单词是否存在bagOfWords2VecMN(vocabList, inputSet)
这个是将句子转成向量的另一种模型，多项式模型，考虑某个词的出现次数trainNB0(trainMatrix,trainCatergory)
计算P(i)和P(w[i]|C[1])和P(w[i]|C[0])，这里有两个技巧，一个是开始的分子分母没有全部初始化为0是为了防止其中一个的概率为0导致整体为0，另一个是后面乘用对数防止因为精度问题结果为0classifyNB(vec2Classify, p0Vec, p1Vec, pClass1)
根据贝叶斯公式计算这个向量属于两个集合中哪个的概率高
1#coding=utf-82fromnumpyimport*3defloadDataSet():4postingList=[['my','dog','has','flea','PRoblems','help','please'],5['maybe','not','take','him','to','dog','park','stupid'],6['my','dalmation','is','so','cute','I','love','him'],7['stop','posting','stupid','worthless','garbage'],8['mr','licks','ate','my','steak','how','to','stop','him'],9['quit','buying','worthless','dog','food','stupid']]10classVec = [0,1,0,1,0,1]#1 is abusive, 0 not11returnpostingList,classVec1213#创建一个带有所有单词的列表14defcreateVocabList(dataSet):15vocabSet =set([])16fordocumentindataSet:17vocabSet = vocabSet |set(document)18returnlist(vocabSet)1920defsetOfWords2Vec(vocabList, inputSet):21retVocabList = [0] *len(vocabList)22forwordininputSet:23ifwordinvocabList:24retVocabList[vocabList.index(word)] = 125else:26print'word',word ,'not in dict'27returnretVocabList2829#另一种模型30defbagOfWords2VecMN(vocabList, inputSet):31returnVec = [0]*len(vocabList)32forwordininputSet:33ifwordinvocabList:34returnVec[vocabList.index(word)] += 135returnreturnVec3637deftrainNB0(trainMatrix,trainCatergory):38numTrainDoc =len(trainMatrix)39numWords =len(trainMatrix[0])40pAbusive = sum(trainCatergory)/float(numTrainDoc)41#防止多个概率的成绩当中的一个为042p0Num =ones(numWords)43p1Num =ones(numWords)44p0Denom = 2.045p1Denom = 2.046foriinrange(numTrainDoc):47iftrainCatergory[i] == 1:48p1Num +=trainMatrix[i]49p1Denom +=sum(trainMatrix[i])50else:51p0Num +=trainMatrix[i]52p0Denom +=sum(trainMatrix[i])53p1Vect = log(p1Num/p1Denom)#处于精度的考虑，否则很可能到限归零54p0Vect = log(p0Num/p0Denom)55returnp0Vect,p1Vect,pAbusive5657defclassifyNB(vec2Classify, p0Vec, p1Vec, pClass1):58p1 = sum(vec2Classify * p1Vec) + log(pClass1)#element-wise mult59p0 = sum(vec2Classify * p0Vec) + log(1.0 -pClass1)60ifp1 >p0:61return162else:63return06465deftestingNB():66listOPosts,listClasses =loadDataSet()67myVocabList =createVocabList(listOPosts)68trainMat=[]69forpostinDocinlistOPosts:70trainMat.append(setOfWords2Vec(myVocabList, postinDoc))71p0V,p1V,pAb =trainNB0(array(trainMat),array(listClasses))72testEntry = ['love','my','dalmation']73thisDoc =array(setOfWords2Vec(myVocabList, testEntry))74printtestEntry,'classified as:',classifyNB(thisDoc,p0V,p1V,pAb)75testEntry = ['stupid','garbage']76thisDoc =array(setOfWords2Vec(myVocabList, testEntry))77printtestEntry,'classified as:',classifyNB(thisDoc,p0V,p1V,pAb)787980defmain():81testingNB()8283if__name__=='__main__':84main()
机器学习笔记索引
来自为知笔记(Wiz)