基于朴素贝叶斯实现垃圾邮件分类，不用sklearn和nltk

可以使用Python中的numpy和pandas库来实现基于朴素贝叶斯的垃圾邮件分类。以下是一个简单的实现代码：

import numpy as np
import pandas as pd

# 读取数据
data = pd.read_csv('spam.csv', encoding='latin-1')
data = data[['v1', 'v2']]
data = data.rename(columns={'v1': 'label', 'v2': 'text'})

# 分割数据集
train_data = data.sample(frac=0.8, random_state=1)
test_data = data.drop(train_data.index)

# 计算先验概率
spam_count = train_data['label'].value_counts()['spam']
ham_count = train_data['label'].value_counts()['ham']
total_count = len(train_data)
p_spam = spam_count / total_count
p_ham = ham_count / total_count

# 计算条件概率
spam_words = []
ham_words = []
for index, row in train_data.iterrows():
    words = row['text'].split()
    if row['label'] == 'spam':
        spam_words += words
    else:
        ham_words += words

spam_word_count = len(spam_words)
ham_word_count = len(ham_words)

spam_word_dict = {}
ham_word_dict = {}

for word in set(spam_words + ham_words):
    spam_word_dict[word] = (spam_words.count(word) + 1) / (spam_word_count + len(set(spam_words + ham_words)))
    ham_word_dict[word] = (ham_words.count(word) + 1) / (ham_word_count + len(set(spam_words + ham_words)))

# 预测
def predict(text):
    words = text.split()
    p_spam_given_text = p_spam
    p_ham_given_text = p_ham
    for word in words:
        if word in spam_word_dict:
            p_spam_given_text *= spam_word_dict[word]
        else:
            p_spam_given_text *= 1 / (spam_word_count + len(set(spam_words + ham_words)))
        if word in ham_word_dict:
            p_ham_given_text *= ham_word_dict[word]
        else:
            p_ham_given_text *= 1 / (ham_word_count + len(set(spam_words + ham_words)))
    if p_spam_given_text > p_ham_given_text:
        return 'spam'
    else:
        return 'ham'

其中，`data`是一个包含标签和文本的数据集，`train_data`和`test_data`是将数据集分割成训练集和测试集的结果。`p_spam`和`p_ham`是先验概率，`spam_word_dict`和`ham_word_dict`是条件概率。`predict`函数可以对新的文本进行分类。

需要注意的是，这只是一个简单的实现，还有很多可以优化的地方。