12.朴素贝叶斯-垃圾邮件分类

1. 读邮件数据集文件，提取邮件本身与标签。

列表

numpy数组

 1 import csv
 2 
 3 
 4 def read_file():
 5     file_path = r'D://PycharmProjects//naive_bayes//data//SMSSpamCollection'
 6     sms = open(file_path, encoding='utf-8')
 7     csv_reader = csv.reader(sms, delimiter='	')
 8     for r in csv_reader:
 9         print(r)
10     sms.close()
11 
12 
13 if __name__ == '__main__':
14     read_file()

2.邮件预处理

邮件分句
句子分词
大小写，标点符号，去掉过短的单词
词性还原：复数、时态、比较级
连接成字符串

2.1 传统方法来实现

 1 '''
 2 传统方法实现
 3 '''
 4 # 利用列表、字典、集合等操作进行词频统计
 5 sep = '.,:;?!-_'
 6 exclude = {'a', 'the', 'and', 'i', 'you', 'in'}
 7 
 8 
 9 def gettxt():
10     txt = open(r'D://PycharmProjects//naive_bayes//data//test.txt', 'r').read().lower()  # 大小写
11     for ch in sep:
12         txt = txt.replace(ch, '')  # 标点符号
13     return txt
14 
15 
16 bigstr = gettxt()  # 获取待统计字符串
17 biglist = bigstr.split()  # 英文分词列表
18 bigdict = {}
19 for word in biglist:
20     bigdict[word] = bigdict.get(word, 0) + 1  # 词频统计字典
21 for word in exclude:
22     del(bigdict[word])  # 无意义词
23 bigitems = list(bigdict.items())
24 bigitems.sort(key=lambda x: x[1], reverse=True)   # 按词频排序
25 for i in range(10):
26     w, c = bigitems[i]
27     print('{0:>10}：{1:<5}'.format(w, c))  # TOP10

2.2 nltk库的安装与使用

pip install nltk

import nltk

nltk.download() # sever地址改成 http://www.nltk.org/nltk_data/

或

https://github.com/nltk/nltk_data下载gh-pages分支，里面的Packages就是我们要的资源。

将Packages文件夹改名为nltk_data。

或

网盘链接：https://pan.baidu.com/s/1iJGCrz4fW3uYpuquB5jbew 提取码：o5ea

放在用户目录。

----------------------------------

安装完成，通过下述命令可查看nltk版本：

import nltk

print nltk.__doc__

1 import nltk
2 
3 print(nltk.__doc__)

2.1 nltk库分词

nltk.sent_tokenize(text) #对文本按照句子进行分割

nltk.word_tokenize(sent) #对句子进行分词

1 import nltk
2 
3 text = "I've been searching for the right words to thank you for this breather. I promise i wont take your help for " 
4        "granted and will fulfil my promise. You have been wonderful and a blessing at all times."
5 
6 sents = nltk.sent_tokenize(text)
7 sents

1 nltk.word_tokenize(sents[0])
2 text.split()

2.2 punkt 停用词

from nltk.corpus import stopwords

stops=stopwords.words('english')

*如果提示需要下载punkt

nltk.download(‘punkt’)

或下载punkt.zip

https://pan.baidu.com/s/1OwLB0O8fBWkdLx8VJ-9uNQ 密码：mema

复制到对应的失败的目录C:UsersAdministratorAppDataRoaming ltk_data okenizers并解压。

1 from nltk.corpus import stopwords
2 
3 stops = stopwords.words('english')
4 print(stops)
5 print(len(stops))

 1 text = "I've been searching for the right words to thank you for this breather. I promise i wont take your help for " 
 2         "granted and will fulfil my promise. You have been wonderful and a blessing at all times."
 3 '''
 4 方法一
 5 '''
 6 # tokens = []
 7 # for sent in nltk.sent_tokenize(text):
 8 #     for word in nltk.word_tokenize(sent):
 9 #         tokens.append(word)
10 # print(tokens)
11 '''
12 方法二
13 '''
14 tokens = [word for sent in nltk.sent_tokenize(text) for word in nltk.word_tokenize(sent)]
15 print("分词后的句子：
", tokens)
16 print("总共有", len(tokens), "个单词")
17 
18 tokens = [token for token in tokens if token not in stops]
19 print("去除停用词后的句子：
", tokens)
20 print("总共有", len(tokens), "个单词")

2.3 NLTK 词性标注

1 nltk.pos_tag(tokens)

2.4 Lemmatisation（词性还原）

from nltk.stem import WordNetLemmatizer

lemmatizer = WordNetLemmatizer()

lemmatizer.lemmatize('leaves') #缺省名词

lemmatizer.lemmatize('best',pos='a')

lemmatizer.lemmatize('made',pos='v')

一般先要分词、词性标注，再按词性做词性还原。

2.5 编写预处理函数

def preprocessing(text):

sms_data.append(preprocessing(line[1])) #对每封邮件做预处理

 1 from nltk.corpus import stopwords
 2 import nltk
 3 import csv
 4 
 5 
 6 def preprocessing(text):
 7     """
 8     预处理
 9     """
10     # text = text.decode("utf-8")
11     tokens = [word for sent in nltk.sent_tokenize(text) for word in nltk.word_tokenize(sent)]  # 分词
12     stops = stopwords.words('english')  # 使用英文的停用词表
13     tokens = [token for token in tokens if token not in stops]  # 去除停用词
14 
15     preprocessed_text = ' '.join(tokens)
16     return preprocessed_text
17 
18 
19 def create_dataset():
20     """
21     导入数据
22     """
23     file_path = r'D://PycharmProjects//naive_bayes//data//SMSSpamCollection'
24     sms = open(file_path, encoding='utf-8')
25     sms_data = []
26     sms_label = []
27     csv_reader = csv.reader(sms, delimiter='	')
28     for line in csv_reader:
29         sms_label.append(line[0])  # 提取出标签
30         sms_data.append(preprocessing(line[1]))  # 提取出特征
31     sms.close()
32     print("数据集标签：
", sms_label)
33     print("数据集特征：
", sms_data)
34 
35 
36 if __name__ == '__main__':
37     create_dataset()

3. 训练集与测试集

4. 词向量

5. 模型

学习链接：https://blog.csdn.net/tinyjian/article/details/79110495