import numpy as np

def loadDataSet():
    postingList=[[‘my‘, ‘dog‘, ‘has‘, ‘flea‘, ‘problems‘, ‘help‘, ‘please‘],    #每一行词表，代表一个文档
                 [‘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]    #标签值    1 is abusive, 0 not
    return postingList,classVec

def createVocabList(dataSet):   #创建词汇列表 将多个文档单词归为一个
    VocabSet = set([]) #使用集合，方便去重
    for document in dataSet:
        VocabSet = VocabSet | set(document)   #保持同类型　　--- 这里并集导致随机序列出现，所以在下面进行排序，方便对比。　　但是单独测试得时候并没有出现随机现象，多次调试，结果唯一，这里每次都不唯一
    VocabList = list(VocabSet)
    VocabList.sort()　　#列表数据排序
    return VocabList

def WordSet2Vec(VocaList,inputSet):  #根据我们上面得到的全部词汇列表，将我们输入得inputSet文档向量化
    returnVec = [0]*len(VocaList)
    for word in inputSet:
        if word in VocaList:
            returnVec[VocaList.index(word)] = 1
        else:
            print("the word: %s is not in my Vocabulary!"%word)
    return returnVec

DocData,classVec = loadDataSet()
voclist = createVocabList(DocData)
print(voclist)
print(WordSet2Vec(voclist,DocData[0]))
print(WordSet2Vec(voclist,DocData[3]))

def trainNB0(trainMatrix,trainCategory):    #训练朴素贝叶斯模型 传入numpy数组类型 trainMatrix所有文档词汇向量矩阵（m*n矩阵 m个文档，每个文档都是n列，代表词汇向量大小),trainCategory每篇文档得标签
    numTrainDoc = len(trainMatrix)  #文档数量
    pC1 = np.sum(trainCategory)/numTrainDoc    #p(c1)得概率   p(c0)=1-p(c1)
    wordVecNum = len(trainMatrix[0])    #因为每个文档转换为词汇向量后都是一样得长度

    #初始化p1,p0概率向量
    p1VecNum,p0VecNum = np.zeros(wordVecNum),np.zeros(wordVecNum)
    p1Sum,p0Sum = 0,0

    #循环每一个文档
    for i in range(numTrainDoc):
        if trainCategory[i] == 1: #侮辱性文档
            p1VecNum += trainMatrix[i]  #统计侮辱性文档中，每个单词出现频率
            p1Sum += np.sum(trainMatrix[i]) #统计侮辱性文档中出现得全部单词数   每个单词出现概率就是单词出现频率/全部单词
        else:   #正常文档
            p0VecNum += trainMatrix[i]
            p0Sum += np.sum(trainMatrix[i])

    p1Vect = p1VecNum / p1Sum   #统计各类文档中的单词出现频率
    p0Vect = p0VecNum / p0Sum

    return p1Vect,p0Vect,pC1

DocData,classVec = loadDataSet()
voclist = createVocabList(DocData)

#获取全部文档词汇向量矩阵
trainMulList = []
for doc in DocData:
    trainMulList.append(WordSet2Vec(voclist,doc))

#将我们需要的词汇列表和标签列表转为numpy.ndarray类型
trainMat = np.array(trainMulList)
trainClassVec = np.array(classVec)

#获取训练模型
p1vec,p0vec,pc1 = trainNB0(trainMat,trainClassVec)

print(p1vec)
print(p0vec)
print(pc1)

def trainNB0(trainMatrix,trainCategory):    #训练朴素贝叶斯模型 传入numpy数组类型 trainMatrix所有文档词汇向量矩阵（m*n矩阵 m个文档，每个文档都是n列，代表词汇向量大小),trainCategory每篇文档得标签
    numTrainDoc = len(trainMatrix)  #文档数量
    pC1 = np.sum(trainCategory)/numTrainDoc    #p(c1)得概率   p(c0)=1-p(c1)
    wordVecNum = len(trainMatrix[0])    #因为每个文档转换为词汇向量后都是一样得长度

    #初始化p1,p0概率向量---改进为拉普拉斯平滑
    p1VecNum,p0VecNum = np.ones(wordVecNum),np.ones(wordVecNum)
    p1Sum,p0Sum = 2.0,2.0   #N*1    N表示分类数

    #循环每一个文档
    for i in range(numTrainDoc):
        if trainCategory[i] == 1: #侮辱性文档
            p1VecNum += trainMatrix[i]  #统计侮辱性文档中，每个单词出现频率
            p1Sum += np.sum(trainMatrix[i]) #统计侮辱性文档中出现得全部单词数   每个单词出现概率就是单词出现频率/全部单词
        else:   #正常文档
            p0VecNum += trainMatrix[i]
            p0Sum += np.sum(trainMatrix[i])

    p1Vect = np.log(p1VecNum / p1Sum)   #统计各类文档中的单词出现频率
    p0Vect = np.log(p0VecNum / p0Sum)   #使用对数避免下溢

    return p1Vect,p0Vect,pC1

def classifyNB(testVec,p0Vec,p1Vec,pC1):
    p1 = sum(testVec*p1Vec)+np.log(pC1) #使用对数之后变为求和
    p0 = sum(testVec*p0Vec)+np.log(1-pC1)
    if p1 > p0:
        return 1
    else:
        return 0

DocData,classVec = loadDataSet()
voclist = createVocabList(DocData)

#获取全部文档词汇向量矩阵
trainMulList = []
for doc in DocData:
    trainMulList.append(WordSet2Vec(voclist,doc))

#将我们需要的词汇列表和标签列表转为numpy.ndarray类型
trainMat = np.array(trainMulList)
trainClassVec = np.array(classVec)

#获取训练模型
p1vec,p0vec,pc1 = trainNB0(trainMat,trainClassVec)

#设置测试集进行测试
testEntry01 = ["love","my","dalmation"]
VecEntry01 = WordSet2Vec(voclist,testEntry01)
clf = classifyNB(VecEntry01,p0vec,p1vec,pc1)
print(clf)

testEntry02 = ["stupid","garbage"]
VecEntry02 = WordSet2Vec(voclist,testEntry02)
clf = classifyNB(VecEntry02,p0vec,p1vec,pc1)
print(clf)