• 博客代码：180927

• 作业代码：181007

• 贝叶斯概率

朴素贝叶斯概率计算过程

• 邮箱分类器

邮件的区分方法

 

• 处理的主要流程

处理的流程

 

• 具体的代码实现

ftrain = open("spam_train.txt")lines = ftrain.readlines()S = dict()      #初始化a,b字典H = dict()tagS = dict()   #标记位，使得一个邮件里面不同的词只会记录一次tagH = dict()trash = 0       #记录垃圾邮箱的个数for line in lines:      #对a，b以及标记位进行初始化，从0开始计数    data = line.split(' ')    count_d = len(data)    if data[0] == '0':        trash+=1    for i in range(count_d-1):        S[data[i+1]] = 0        tagS[data[i+1]] = 0        H[data[i+1]] = 0        tagH[data[i+1]] = 0for line in lines:      #记录关键词垃圾邮件和正常邮件的个数    for i in range(count_d-1):        tagS[data[i+1]] = 0        tagH[data[i+1]] = 0            data = line.split(' ')    count_d = len(data)    if data[0] == '0':        for i in range(count_d-1):            if tagS[data[i+1]] == 0:    #如果没被标记过，才会对其进行加数                S[data[i+1]]+=1                tagS[data[i+1]] = 1                                if data[0] == '1':         for i in range(count_d-1):            if tagH[data[i+1]] == 0:                H[data[i+1]]+=1                tagH[data[i+1]] = 1    rig = 0ftest = open("spam_test.txt")lines = ftest.readlines()for line in lines:    ans = trash/(5000-trash)    data = line.split(' ')    count_d = len(data)    for i in range(count_d - 1):        #将没有出现在训练邮件的单词置0        S.setdefault(data[i+1],0)    for i in range(count_d - 1):        H.setdefault(data[i+1],0)    for i in range(count_d - 1):        #将两个结果进行相除比较大小        ans = ans*(S.get(data[i+1])+1)/(trash + 1)        ans = ans*(5001 - trash)/(H.get(data[i+1])+1)    if ans>1:        if data[0] == '0':            rig+=1    if ans<1:        if data[0] == '1':            rig+=1print(rig/1000)