naiveBayes_python

#coding:utf-8
from numpy import *
import re
def createlist(lst):#将所有文本放入一个列表中
    listt=set([])
    for line in lst:
        listt=listt|set(line)
    return list(listt)
def word2vec(List,inputset):#将输入文本转为词向量,每个文本对应一个词向量,其长度为上述列表长度
    lenth=len(List)
    vec=[0]*lenth
    for word in inputset:
        if word in List:
            vec[List.index(word)]+=1
    return vec
def singprob(trainmatrix,label):#计算在已知类别的条件下,每个单词出现的概率对应于p(w1|c1),p(w2|c1).....
    lenth=len(trainmatrix)
    numword=len(trainmatrix[0])
    pb1=sum(label)/float(lenth)
    p0num=ones(numword)
    p1num=ones(numword)
    p0all=2
    p1all=2
    for i in range(lenth):
        if label[i]==1:
            p1num+=trainmatrix[i]
            p1all+=sum(trainmatrix[i])
        else:
            p0num+=trainmatrix[i]
            p0all+=sum(trainmatrix[i])
    p1vect=log(p1num/p1all)
    p0vect=log(p0num/p0all)
    return p1vect,p0vect,pb1
def classifier(vect,p1vect,p0vect,pb1):#对应公式:lnp(w1|c=1 or 0)p(w2/c=1 or 0)..p(wn/c=1 0r 0)p(c)
    p1=sum(vect*p1vect)+log(pb1)
    p0=sum(vect*p0vect)+log(1-pb1)
    if p1>p0:
        return 1
    else:
        return 0
def testparse(str):
    reg=re.compile('\W*')
    line=reg.split(str)
    List=[tt.lower for tt in line if len(tt)>2]
    return List
def Test():
    doc=[]
    label=[]
    fulltext=[]
    for i in range(1,26):
        wordlist=testparse(open("email/spam/%d.txt" %i).read())
        doc.append(wordlist)
        fulltext.extend(wordlist)
        label.append(1)
        wordlist=testparse(open("email/ham/%d.txt" %i).read())
        doc.append(wordlist)
        fulltext.extend(wordlist)
        label.append(0)
    doclist=createlist(doc)
    trainingset=range(50)
    testset=[]
    for i in range(10):
        index=int(random.uniform(0,len(trainingset)))
        testset.append(index)
        del(trainingset[index])
    trainmat=[]
    classlabel=[]
    for docindex in trainingset:
        trainmat.append(word2vec(doclist,doc[docindex]))
        classlabel.append(label[docindex])
    p1,p0,pb=singprob(trainmat,classlabel)
    error=0
    for testindex in testset:
        wordvect=word2vec(doclist,doc[testindex])
        if classifier(wordvect,p1,p0,pb)!=label[testindex]:
            error+=1
    print error/float(len(testset))
Test()