利用朴素贝叶斯公式和词袋模型进行 垃圾邮件的判断实战

案例说明

垃圾邮件是最基础的机器学习案例，其中涉及到的知识也比较全面。所以作为一个案例来进行学习和总结

目标：

将垃圾邮件自动识别， 其实是个二分类问题，我们需要将目标进行分解， 降低实现的难度。

步骤

1、进行数据的读取，把数据划分为垃圾和正常两个类型

def open_file(path):
    list = []
    #读取文件
    with open(path, encoding='utf-8') as lines:
        for line in lines:
            line = line.strip('\n')
            #因为词袋模型是用一个词语来表示频率，所以需要用jieba库进行分词，得到的是迭代器generator，可以用join的方法转回字符串。
            seg_list = jieba.cut(clean_str(line), cut_all=False)
            str = " ".join(seg_list)
            # 放在一个数组里，存放的格式为：
            ['wo 是 大 大', ...]
            list.append(str)
    return list

2、将内容进行词频化，正常的邮件内容是文本，我们需要将其转换成数字，借助sklearn的
from sklearn.feature_extraction.text import CountVectorizer
完成

def format_data():
    # 返回数据数组
    x_pos = open_file('./data/ham_5000.utf8')
    x_neg = open_file('./data/spam_5000.utf8')
    # 生成标签集
    y_pos = [1] * len(x_pos)
    y_neg = [0] * len(x_neg)
    # 数据集整合
    X = x_pos + x_neg
    y = y_pos + y_neg
    # 划分数据集
    x_train, x_test, y_train, y_test = train_test_split(X, y, test_size=0.3)
    # 向量化数据集 先统计词频， 再逆向带入变成向量
    counter_vec = CountVectorizer(max_features=1000)
    x_train = counter_vec.fit_transform(x_train)
    x_test = counter_vec.transform(x_test)
    return x_train, x_test, y_train, y_test

3、接着将数据传入模型进行训练，这里用的是最简单的平滑MultinomialNB模型，不需要参数的定义。

def main():
    x_train, x_test, y_train, y_test = format_data()
    print(x_train)
    model = MultinomialNB()
    model.fit(x_train, y_train)
    y_pre = model.predict(x_test)
    r = classification_report(y_test, y_pre)
    print(r)

可以看到准确率和recall还是比较高的，达到了97%

数据链接：点击

完整代码：

from sklearn.feature_extraction.text import CountVectorizer
import numpy as np
import jieba
from sklearn.model_selection import train_test_split
from sklearn.naive_bayes import MultinomialNB
from sklearn.metrics import classification_report
import re
def clean_str(s):
    return  re.sub('[\.\。\，\“\”\！\？\!]', '', s)
def open_file(path):
    list = []
    with open(path, encoding='utf-8') as lines:
        for line in lines:
            line = line.strip('\n')
            seg_list = jieba.cut(clean_str(line), cut_all=False)
            str = " ".join(seg_list)
            list.append(str)
    return list
def format_data():
    # 返回数据数组
    x_pos = open_file('./data/ham_5000.utf8')
    x_neg = open_file('./data/spam_5000.utf8')
    # 生成标签集
    y_pos = [1] * len(x_pos)
    y_neg = [0] * len(x_neg)
    # 数据集整合
    X = x_pos + x_neg
    y = y_pos + y_neg
    # 划分数据集
    x_train, x_test, y_train, y_test = train_test_split(X, y, test_size=0.3)
    # 向量化数据集 先统计词频， 再逆向带入变成向量
    counter_vec = CountVectorizer(max_features=1000)
    x_train = counter_vec.fit_transform(x_train)
    x_test = counter_vec.transform(x_test)
    return x_train, x_test, y_train, y_test
def main():
    x_train, x_test, y_train, y_test = format_data()
    print(x_train)
    model = MultinomialNB()
    model.fit(x_train, y_train)
    y_pre = model.predict(x_test)
    r = classification_report(y_test, y_pre)
    print(r)
main()