📖 第14课：垃圾邮件检测

from sklearn.feature_extraction.text import TfidfVectorizer from sklearn.naive_bayes import MultinomialNB from sklearn.metrics import classification_report, confusion_matrix import numpy as np # 模拟邮件数据集 emails = [ "恭喜您中奖了，点击领取百万大奖", "限时优惠，折扣低至一折，立即抢购", "免费领取iPhone，数量有限先到先得", "赚钱秘籍，月入百万不是梦", "尊敬的用户，您的账户存在异常，请立即验证", "双十一特惠，全场半价起", "投资理财回报率高达300%", "恭喜被选中参加幸运抽奖", "张总，明天的会议改到下午三点", "附件是本月的销售报告，请查收", "项目进度更新：前端开发已完成80%", "关于下周出差安排的通知", "请审核一下合同条款", "团队聚餐安排：周五晚上七点", "系统维护通知：本周六凌晨进行升级", "季度工作总结已提交", "客户反馈汇总，请尽快回复", "新员工入职培训安排", ] labels = ['spam'] * 8 + ['ham'] * 10 vectorizer = TfidfVectorizer(token_pattern=r'(?u)\b\w+\b') X = vectorizer.fit_transform(emails) clf = MultinomialNB(alpha=0.5) clf.fit(X, labels) # 预测新邮件 test_emails = [ "恭喜中奖！点击领取", "请查看附件中的报告", "限时免费领取大奖", "会议安排通知", "投资理财高回报", "项目进度更新" ] test_X = vectorizer.transform(test_emails) preds = clf.predict(test_X) probs = clf.predict_proba(test_X) print("=== 垃圾邮件检测 ===") for email, pred, prob in zip(test_emails, preds, probs): idx = list(clf.classes_).index(pred) print(f" '{email}' → {pred} (置信度: {prob[idx]:.3f})") # 特征分析 print("\n=== 垃圾邮件关键词 ===") feature_names = vectorizer.get_feature_names_out() spam_probs = clf.feature_log_prob_[list(clf.classes_).index('spam')] top_spam_idx = np.argsort(spam_probs)[-10:][::-1] for idx in top_spam_idx: print(f" {feature_names[idx]}: log_prob={spam_probs[idx]:.4f}")

import re class RuleBasedSpamFilter: def __init__(self): self.spam_patterns = [ r'中奖', r'免费领取', r'月入\d+万', r'折扣.*低至', r'恭喜.*被选', r'高回报', r'限时.*优惠', r'投资.*理财.*\d+%', r'点击.*领取', r'账号.*异常.*验证', ] self.spam_words = {'大奖', '抢购', '特惠', '中奖', '免费', '赚钱', '秘籍', '优惠'} def score(self, text): score = 0.0 for pattern in self.spam_patterns: if re.search(pattern, text): score += 0.3 for word in self.spam_words: if word in text: score += 0.2 return min(score, 1.0) def classify(self, text, threshold=0.4): score = self.score(text) return 'spam' if score >= threshold else 'ham', score filter = RuleBasedSpamFilter() test_emails = [ "恭喜中奖！点击领取百万大奖", "明天的会议改到下午", "限时优惠，免费领取iPhone", "项目进度报告请查收", "投资理财高回报300%", "合同审核通知", ] print("=== 规则型垃圾邮件过滤器 ===") for email in test_emails: label, score = filter.classify(email) bar = '█' * int(score * 20) + '░' * (20 - int(score * 20)) print(f" [{bar}] {score:.2f} {label:4s} | {email}")