機器學習之貝葉斯垃圾郵件分類

代碼來源於：https://www.cnblogs.com/huangyc/p/10327209.html  ，本人只是簡介學習

1、 貝葉斯.py

import numpy as np
from word_utils import *



class NaiveBayesBase(object):

    def __init__(self):
        pass


    def fit(self, trainMatrix, trainCategory):
        '''
        朴素貝葉斯分類器訓練函數，求：p(Ci),基於詞匯表的p(w|Ci)
        Args:
            trainMatrix : 訓練矩陣，即向量化表示后的文檔（詞條集合）
            trainCategory : 文檔中每個詞條的列表標注
        Return:
            p0Vect : 屬於0類別的概率向量(p(w1|C0),p(w2|C0),...,p(wn|C0))
            p1Vect : 屬於1類別的概率向量(p(w1|C1),p(w2|C1),...,p(wn|C1))
            pAbusive : 屬於1類別文檔的概率
        '''
        numTrainDocs = len(trainMatrix)
        # 長度為詞匯表長度
        numWords = len(trainMatrix[0])
        # p(ci)
        self.pAbusive = sum(trainCategory) / float(numTrainDocs)
        # 由於后期要計算p(w|Ci)=p(w1|Ci)*p(w2|Ci)*...*p(wn|Ci)，若wj未出現，則p(wj|Ci)=0,因此p(w|Ci)=0，這樣顯然是不對的
        # 故在初始化時，將所有詞的出現數初始化為1，分母即出現詞條總數初始化為2
        p0Num = np.ones(numWords)
        p1Num = np.ones(numWords)
        p0Denom = 2.0
        p1Denom = 2.0
        for i in range(numTrainDocs):
            if trainCategory[i] == 1:
                p1Num += trainMatrix[i]
                p1Denom += sum(trainMatrix[i])
            else:
                p0Num += trainMatrix[i]
                p0Denom += sum(trainMatrix[i])
        # p(wi | c1)
        # 為了避免下溢出（當所有的p都很小時，再相乘會得到0.0，使用log則會避免得到0.0）
        self.p1Vect = np.log(p1Num / p1Denom)
        # p(wi | c2)
        self.p0Vect = np.log(p0Num / p0Denom)
        return self


    def predict(self, testX):
        '''
        朴素貝葉斯分類器
        Args:
            testX : 待分類的文檔向量（已轉換成array）
            p0Vect : p(w|C0)
            p1Vect : p(w|C1)
            pAbusive : p(C1)
        Return:
            1 : 為侮辱性文檔 (基於當前文檔的p(w|C1)*p(C1)=log(基於當前文檔的p(w|C1))+log(p(C1)))
            0 : 非侮辱性文檔 (基於當前文檔的p(w|C0)*p(C0)=log(基於當前文檔的p(w|C0))+log(p(C0)))
        '''

        p1 = np.sum(testX * self.p1Vect) + np.log(self.pAbusive)
        p0 = np.sum(testX * self.p0Vect) + np.log(1 - self.pAbusive)
        if p1 > p0:
            return 1
        else:
            return 0

def loadDataSet():
    '''數據加載函數。這里是一個小例子'''
    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]  # 1代表侮辱性文字，0代表正常言論，代表上面6個樣本的類別
    return postingList, classVec


def checkNB():
    '''測試'''
    listPosts, listClasses = loadDataSet()
    myVocabList = createVocabList(listPosts)
    trainMat = []
    for postDoc in listPosts:
        trainMat.append(setOfWord2Vec(myVocabList, postDoc))

    nb = NaiveBayesBase()
    nb.fit(np.array(trainMat), np.array(listClasses))

    testEntry1 = ['love', 'my', 'dalmation']
    thisDoc = np.array(setOfWord2Vec(myVocabList, testEntry1))
    print(testEntry1, 'classified as:', nb.predict(thisDoc))

    testEntry2 = ['stupid', 'garbage']
    thisDoc2 = np.array(setOfWord2Vec(myVocabList, testEntry2))
    print(testEntry2, 'classified as:', nb.predict(thisDoc2))


if __name__ == "__main__":
    checkNB()

2、word_utils.py

def createVocabList(dataSet):
    '''
    創建所有文檔中出現的不重復詞匯列表
    Args:
        dataSet: 所有文檔
    Return:
        包含所有文檔的不重復詞列表，即詞匯表
    '''
    vocabSet = set([])
    # 創建兩個集合的並集
    for document in dataSet:
        vocabSet = vocabSet | set(document)
    return list(vocabSet)


# 詞袋模型(bag-of-words model):詞在文檔中出現的次數
def bagOfWords2Vec(vocabList, inputSet):
    '''
    依據詞匯表，將輸入文本轉化成詞袋模型詞向量
    Args:
        vocabList: 詞匯表
        inputSet: 當前輸入文檔
    Return:
        returnVec: 轉換成詞向量的文檔
    例子：
        vocabList = ['I', 'love', 'python', 'and', 'machine', 'learning']
        inputset = ['python', 'machine', 'learning', 'python', 'machine']
        returnVec = [0, 0, 2, 0, 2, 1]
        長度與詞匯表一樣長，出現了的位置為1，未出現為0，如果詞匯表中無該單詞則print
    '''
    returnVec = [0] * len(vocabList)
    for word in inputSet:
        if word in vocabList:
            returnVec[vocabList.index(word)] += 1
        else:
            print("the word: %s is not in my vocabulary!" % word)
        return returnVec


# 詞集模型(set-of-words model):詞在文檔中是否存在，存在為1，不存在為0
def setOfWord2Vec(vocabList, inputSet):
    '''
    依據詞匯表，將輸入文本轉化成詞集模型詞向量
    Args:
        vocabList: 詞匯表
        inputSet: 當前輸入文檔
    Return:
        returnVec: 轉換成詞向量的文檔
    例子：
        vocabList = ['I', 'love', 'python', 'and', 'machine', 'learning']
        inputset = ['python', 'machine', 'learning']
        returnVec = [0, 0, 1, 0, 1, 1]
        長度與詞匯表一樣長，出現了的位置為1，未出現為0，如果詞匯表中無該單詞則print
    '''
    returnVec = [0] * len(vocabList)
    for word in inputSet:
        if word in vocabList:
            returnVec[vocabList.index(word)] = 1
        else:
            print("the word: %s is not in my vocabulary!" % word)
    return returnVec