13 垃圾邮件分类2

读取

1 # 读取文件
2 sms = open(r'D:\机器学习\垃圾邮件分类\SMSSpamCollection','r',encoding = 'utf-8')
3 sms_label = []  # 字符串的列表
4 sms_data = []   # 数据
5 csv_reader = csv.reader(sms, delimiter = '\t')
6 for line in csv_reader:
7     sms_label.append(line[0])
8     sms_data.append(preprocessing(line[1])) # 对每封邮件做预处理，生成有效词的字符串
9 sms.close()
数据预处理

1 # 定义词性还原的函数
 2 def get_wordnet_pos(treebank_tag):  # 还原参数pos
 3     if treebank_tag.startswith('J'):
 4         return nltk.corpus.wordnet.ADJ
 5     elif treebank_tag.startswith('V'):
 6         return nltk.corpus.wordnet.VERB
 7     elif treebank_tag.startswith('N'):
 8         return nltk.corpus.wordnet.NOUN
 9     elif treebank_tag.startswith('R'):
10         return nltk.corpus.wordnet.ADV
11     else:
12         return nltk.corpus.wordnet.NOUN
13 
14 # 预处理
15 def preprocessing(text):
16     tokens = [word for sent in nltk.sent_tokenize(text) for word in nltk.word_tokenize(sent)]   # 分词
17     stops = stopwords.words('english')  # 停用词类型为英文
18     tokens = [token for token in tokens if token not in stops]  # 保留非停用词
19     tokens = [token.lower() for token in tokens if len(token)>=3]   # 大写转小写,去掉长度短于3的词
20 
21     tag = nltk.pos_tag(tokens)  # 标注词性
22     lemmatizer = WordNetLemmatizer()  # 词性还原
23     tokens = [lemmatizer.lemmatize(token, pos = get_wordnet_pos(tag[i][1])) for i, token in enumerate(tokens)] # 词性还原
24     preprocessed_text = ''.join(tokens)
25     return preprocessed_text
数据划分—训练集和测试集数据划分

from sklearn.model_selection import train_test_split

x_train,x_test, y_train, y_test = train_test_split(data, target, test_size=0.2, random_state=0, stratify=y_train)

代码：

1 # 划分训练集跟测试集
2 x_train, x_test, y_train, y_test = train_test_split(sms_data, sms_label, test_size = 0.2, stratify = sms_label) # 测试集占20%
3 print('总数据量：',len(sms_label))
4 print('训练数据量：',len(x_train))
5 print('测试数据量：',len(y_test))
文本特征提取

sklearn.feature_extraction.text.CountVectorizer

https://scikit-learn.org/stable/modules/generated/sklearn.feature_extraction.text.CountVectorizer.html?highlight=sklearn%20feature_extraction%20text%20tfidfvectorizer

sklearn.feature_extraction.text.TfidfVectorizer

https://scikit-learn.org/stable/modules/generated/sklearn.feature_extraction.text.TfidfVectorizer.html?highlight=sklearn%20feature_extraction%20text%20tfidfvectorizer#sklearn.feature_extraction.text.TfidfVectorizer

from sklearn.feature_extraction.text import TfidfVectorizer

tfidf2 = TfidfVectorizer()

1 # 向量化
2 tfidf2 = TfidfVectorizer()
3 X_train = tfidf2.fit_transform(x_train)
4 X_test = tfidf2.transform(x_test)
5 print(X_train.toarray().shape)
6 print(X_test.toarray().shape)
模型选择

from sklearn.naive_bayes import GaussianNB

from sklearn.naive_bayes import MultinomialNB

说明为什么选择这个模型？

1 mnb = MultinomialNB()
2 mnb.fit(X_train, y_train)
3 y_mnb = mnb.predict(X_test)
4 print('预测值：',y_mnb)
5 print('实际值：',y_test)
模型评价：混淆矩阵，分类报告

from sklearn.metrics import confusion_matrix

confusion_matrix = confusion_matrix(y_test, y_predict)

说明混淆矩阵的含义

from sklearn.metrics import classification_report

说明准确率、精确率、召回率、F值分别代表的意义

1 # 混淆矩阵，分类报告
2 cm = confusion_matrix(y_test, y_mnb)
3 print(cm)
4 cr = classification_report(y_test, y_mnb)
5 print(cr)

 

 比较与总结

如果用CountVectorizer进行文本特征生成，与TfidfVectorizer相比，效果如何？

答：效果较差，因为用CountVectorizer进行文本特征生成，会出现正常邮件被划分为垃圾邮件的情况