基于朴素贝叶斯分类的恶意言论分类简单测试

若结果有误，请指出噢

#1找到所有特征
import numpy as np
#from numpy import *
def loadDataSet():
    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]#标签1是恶评，0不是恶评
    return postingList,classVec
#listOPosts,listclasses = loadDataSet()
#print(listOPosts)
def createVocaList(dataSet):
    vocabSet = set([])#创建空的set集合，无序不重复元素集
    for document in dataSet:
        vocabSet = vocabSet|set(document)#|代表并集，并集里面没有重复元素
    return list(vocabSet)#列表，因为set集合没有索引

#MyvocabList = createVocaList(listOPosts)
#print('特征属性：',MyvocabList)

# 2、找到每条言论对应的词汇向量，每一条言论对应的词汇量是一个列表的形式
def setofwords2Vec(feature,inputset):
    returnVec = [0]*len(feature)
    for word in inputset:
        if word in feature:
            returnVec[feature.index(word)] = 1
        else:
             print('the word:%s is not in my vocabList'%word)
    return returnVec
#trainMatrix = []
# for postinDoc in listOPosts:
#        print(postinDoc)
#        trainMatrix.append(setofwords2Vec(MyvocabList,postinDoc))
#        print(trainMatrix)

#3、根据词汇，求出来每个特征在每种标签下对应的概率
def trainNB0(wordSetVec,labels):
    numDataSet = len(wordSetVec)#样本数
    pAbusive = sum(labels)/numDataSet#恶意言论的概率
    featureNum = len(wordSetVec[0])#特征数
    #初始化零（行）向量，存放每个特征在每种标签下出现的次数，0（1）表示善意（恶意）言论
    p0Num = np.ones(featureNum);p1Num = np.ones(featureNum)
    #初始化分母，分母代表的是每种标签里面的单词数
    p0Denom = 2;p1Denom = 2
    for i in range(numDataSet):
        if labels[i] == 1:
            p1Num += wordSetVec[i]#对应位置上相加。累加即可求得每个特征出现的次数
            p1Denom += sum(wordSetVec[i])#累加即可求得1标签下所有单词数
        else:
            p0Num += wordSetVec[i]#对应位置上相加。累加即可求得每个特征出现的次数
            p0Denom += sum(wordSetVec[i])#累加即可求得0标签下所有单词数
    p0Vec = np.log(p0Num/p0Denom)#求出每个特征在0类别下的概率
    p1Vec = np.log(p1Num/p1Denom)#求出每个特征在1类别下的概率
    return p0Vec,p1Vec,pAbusive
# p0,p1,pA = trainNB0(trainMatrix,listclasses)
# print('p0:',p0)
#print('p1:',p1)
            

#4、由上面的概率值，定义朴素贝叶斯分类

def NBC(TestWordsVec,p0Vec,p1Vec,pclasses):
    p1 = sum(TestWordsVec*p1Vec) + np.log(pclasses)
    p0 = sum(TestWordsVec*p0Vec) + np.log(1-pclasses)
    if p1>p0:
        return "1标签"
    if p1<p0:
        return "0标签"
#测试算法
def testingNBC():
    listOPosts,listclasses = loadDataSet()
    print('词汇列表：',listOPosts)
    MyvocabList = createVocaList(listOPosts)
    print('特征属性：',MyvocabList)
    trainMatrix = []
    for postinDoc in listOPosts:
#        print(postinDoc)
        trainMatrix.append(setofwords2Vec(MyvocabList,postinDoc))
    print('特征向量：',trainMatrix)
    p0,p1,pA = trainNB0(trainMatrix,listclasses)
    print('概率：',p0,p1,pA)
    testEntry = ['love','my','dalmation']
    testVec = setofwords2Vec(MyvocabList,testEntry)
    print('分类器分类的类别为：',NBC(testVec,p0,p1,pA))
    testEntry1 = ['stupid','garbage']
    testVec1 = setofwords2Vec(MyvocabList,testEntry1)
    print('分类器分类的类别：',NBC(testVec1,p0,p1,pA))
    
testingNBC()