今天看了朴素贝叶斯，在网上看到了篇朴素贝叶斯的文章感觉特别好，在此摘抄 算法杂货铺——分类算法之朴素贝叶斯分类(Naive Bayesian classification)
注：1、
p(ci|w)=p(w|ci)p(ci)p(w) 我们将使用上述公式，对每个计算该值，然后比较这两个概率值大小。 2、利用贝叶斯分类器对文档进行分类时，要计算多个概率的乘积以获得文档属于某个文档类别的概率，即计算 p(w0|1)p(w1|1)p(w2|1) ,如果其中一个概率值为0，那么最后的成绩也为0.为了降低这种影响可以将所有词的出现数初始化为1，并将分母初始化为2 3、另一个遇到的问题是下溢出，这是由于太多很小的数相乘造成的。当计算 p(w0|ci)p(w1|ci)p(w2|ci)……p(wn|ci) ,由于大部分因子都很小，所以程序会下溢出或者得不到正确答案 为了解决这个问题于是对乘积取自然对数，以此避免下溢出或者浮点数舍入导致的错误 #encoding:utf-8 import numpy as np from numpy import * def loadDataSet():#创建一个实验样本 postingList=[['my', 'dog', 'has', 'flea', 'problems', 'help', 'please'], ['maybe', 'not', 'take', 'him', 'to', 'dog', 'park', 'stupid'], ['my', 'dalmation', 'is', 'so', 'cute', 'I', 'love', 'him'], ['stop', 'posting', 'stupid', 'worthless', 'garbage'], ['mr', 'licks', 'ate', 'my', 'steak', 'how', ' to', 'stop', 'him'], ['quit', 'buying', 'worthless', 'dog', 'food', 'stupid']] classVec = [0,1,0,1,0,1] return postingList,classVec def creatVocabList(dataSet):#创建一个包含所有文档中出现不重复的词表 vocabSet = set() for document in dataSet: vocabSet = vocabSet | set(document)#集合的合并 return list(vocabSet) def setOfWordsVec(vocabList, inputSet):#输入文档表输出文档向量 returnVec = [0]*len(vocabList)#创建一个行向量 for word in inputSet: if word in vocabList:#如果在词表中出现，则对应词表中的词中的位置变为1 returnVec[vocabList.index(word)] = 1 else: print "the word : %s is not in my Vocabulary!" % word return returnVec#返回的是行向量 def trainNB0(trainMatrix, trainCategory):#朴素贝叶斯分类器的训练函数 numTrainDocs = len(trainMatrix)#样本个数（行向量） numWords = len(trainMatrix[0])#词汇数目（列向量） pAbsive = sum(trainCategory)/float(numTrainDocs)#样本是侮辱性话语的概率 # p0Num = zeros(numWords) # p1Num = zeros(numWords) # p0Denom = 0.0;p1Denom = 0.0 p0Num = ones(numWords) p1Num = ones(numWords) p0Denom = 2.0;p1Denom = 2.0 for i in range(numTrainDocs):#遍历所有样本 if trainCategory[i] == 1:#如果是侮辱性的 p1Num += trainMatrix[i]#统计各个词汇出现的频数 p1Denom += sum(trainMatrix[i])#计算词汇总数 else: p0Num += trainMatrix[i] p0Denom += sum(trainMatrix[i]) # p1Vec = p1Num/float(p1Denom) # p0Vec = p0Num/float(p0Denom) p1Vec = log(p1Num/float(p1Denom)) p0Vec = log(p0Num/float(p0Denom)) return p0Vec, p1Vec, pAbsive#返回p(wi|c1),p(wi|c0),p(c1) wi 词汇表里面的各个词汇c1代表侮辱类 c非侮辱类 def classifyNB(vec2Classify, p0Vec, p1Vec, pClass):#分类器 #vec2Classify 这里起筛选的作用,选出词条的特征 p1 = sum(vec2Classify*p1Vec)+log(pClass)#得出p(w0|1)p(w1|1)p(w2|1)p(1) p0 = sum(vec2Classify*p0Vec)+log(1.0-pClass)#得出p(w0|0)p(w1|0)p(w2|0)p(0) if p1 > p0: return 1 else: return 0 def testingNB(): #测试函数 listOPosts,listClasses = loadDataSet() myVocabList = creatVocabList(listOPosts) trainMat = [] for i in range(len(listOPosts)): trainMat.append(setOfWordsVec(myVocabList, listOPosts[i])) p0V,p1V,pAb = trainNB0(trainMat, listClasses) test = ['love', 'my', 'dalmation'] thisDoc = array(setOfWordsVec(myVocabList, test)) print classifyNB(thisDoc, p0V,p1V,pAb) test = ['stupid', 'garbage'] thisDoc = array(setOfWordsVec(myVocabList, test)) print classifyNB(thisDoc, p0V,p1V,pAb) def textParse(bigString):#文本解析 import re listOfTokens = re.split(r'\W*', bigString)#正则表达 剔除非英文非数字 # print listOfTokens return [tok.lower() for tok in listOfTokens if len(tok) > 2]#只要2个字母以上的单词 def spamTest():#测试函数 docList=[];classList=[];fullText=[] for i in range(1,26):#根据邮件类型建立词表 wordList =textParse(open(r'spam/%d.txt' % i).read())#正常邮件 docList.append(wordList) fullText.append(wordList) classList.append(1) wordList =textParse(open(r'ham/%d.txt' % i).read())#垃圾邮件 docList.append(wordList) fullText.append(wordList) classList.append(0) vocabList = creatVocabList(docList)#建立词汇表 print "-------------" trainSet=range(50);testSet=[] for i in range(10): randIndex = int(random.uniform(0, len(trainSet))) testSet.append(trainSet[randIndex]) del trainSet[randIndex] trainMat=[];trainClass=[] for docIndex in trainSet: trainMat.append(setOfWordsVec(vocabList, docList[docIndex])) trainClass.append(classList[docIndex]) p0V,p1V,pSpam = trainNB0(trainMat, trainClass)#训练 errorCount = 0 for docIndex in testSet:#计算错误率 wordVector = setOfWordsVec(vocabList, docList[docIndex]) if classifyNB(array(wordVector), p0V, p1V, pSpam) != classList[docIndex]: errorCount+=1 print "the error rate is:",float(errorCount)/len(testSet) i = 1 # wordList =textParse(open(r'spam/%d.txt' % i).read()) # print wordList spamTest() # testingNB()