实现朴素贝叶斯分类器（采用拉普拉斯修正）过滤垃圾邮件代码

好的，以下是一个简单的实现朴素贝叶斯分类器过滤垃圾邮件的 Python 代码（采用拉普拉斯修正）：

import os
import re
import math

# 定义垃圾邮件和正常邮件的路径
spam_path = 'spam/'
ham_path = 'ham/'

# 定义停用词列表
stop_words = ['a', 'an', 'the', 'in', 'on', 'at', 'to', 'of', 'for', 'by', 'with', 'from', 'and', 'or', 'not']

# 定义模型的参数
p_spam = 0.0
p_ham = 0.0
p_word_given_spam = {}
p_word_given_ham = {}

# 定义函数：获取文件内容
def get_file_content(file_path):
    with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:
        content = f.read()
    return content

# 定义函数：获取单词列表
def get_word_list(content):
    # 去除标点符号和数字
    content = re.sub(r'[^\w\s]|\d', ' ', content)
    # 转为小写
    content = content.lower()
    # 分词
    word_list = content.split()
    # 去除停用词
    word_list = [word for word in word_list if word not in stop_words]
    return word_list

# 训练模型
def train():
    global p_spam, p_ham, p_word_given_spam, p_word_given_ham
    
    # 统计垃圾邮件和正常邮件的数量
    spam_count = len(os.listdir(spam_path))
    ham_count = len(os.listdir(ham_path))
    
    # 计算先验概率
    p_spam = spam_count / (spam_count + ham_count)
    p_ham = ham_count / (spam_count + ham_count)
    
    # 统计单词数量
    word_count_given_spam = {}
    word_count_given_ham = {}
    for file_name in os.listdir(spam_path):
        file_path = spam_path + file_name
        content = get_file_content(file_path)
        word_list = get_word_list(content)
        for word in word_list:
            word_count_given_spam[word] = word_count_given_spam.get(word, 0) + 1
    for file_name in os.listdir(ham_path):
        file_path = ham_path + file_name
        content = get_file_content(file_path)
        word_list = get_word_list(content)
        for word in word_list:
            word_count_given_ham[word] = word_count_given_ham.get(word, 0) + 1
    
    # 计算条件概率
    for word in word_count_given_spam.keys():
        p_word_given_spam[word] = (word_count_given_spam[word] + 1) / (sum(word_count_given_spam.values()) + len(word_count_given_spam))
    for word in word_count_given_ham.keys():
        p_word_given_ham[word] = (word_count_given_ham[word] + 1) / (sum(word_count_given_ham.values()) + len(word_count_given_ham))

# 预测邮件类型
def predict(file_path):
    content = get_file_content(file_path)
    word_list = get_word_list(content)
    
    # 初始化概率
    p_spam_given_words = math.log(p_spam)
    p_ham_given_words = math.log(p_ham)
    
    # 计算条件概率的对数
    for word in word_list:
        if word in p_word_given_spam:
            p_spam_given_words += math.log(p_word_given_spam[word])
        else:
            p_spam_given_words += math.log(1 / (sum(p_word_given_spam.values()) + len(p_word_given_spam)))
        if word in p_word_given_ham:
            p_ham_given_words += math.log(p_word_given_ham[word])
        else:
            p_ham_given_words += math.log(1 / (sum(p_word_given_ham.values()) + len(p_word_given_ham)))
    
    # 判断邮件类型
    if p_spam_given_words > p_ham_given_words:
        return 'spam'
    else:
        return 'ham'

# 训练模型
train()

# 测试模型
result = predict('spam/1.txt')
print(result)

result = predict('ham/1.txt')
print(result)

其中，`spam/` 和 `ham/` 分别为存放垃圾邮件和正常邮件的文件夹路径，`stop_words` 是停用词列表，`train()` 函数用于训练模型，`predict(file_path)` 函数用于预测邮件类型。在训练模型时，采用了拉普拉斯修正，避免了概率为 0 的情况。在预测邮件类型时，采用了对数概率，防止概率连乘导致下溢。