1.读取
# 1、读取数据集
def read_dataset():
file_path = r'SMSSpamCollection'
sms = open(file_path, encoding='utf-8')
sms_data = []
sms_label = []
csv_reader = csv.reader(sms, delimiter='\t')
for line in csv_reader:
sms_label.append(line[0]) # 提取出标签
sms_data.append(preprocessing(line[1])) # 提取出特征
sms.close()
return sms_data, sms_label2.数据预处理
# 2、数据预处理
def preprocess(text):
tokens = [word for sent in nltk.sent_tokenize(text) for word in nltk.word_tokenize(sent)] # 分词
stops = stopwords.words('english') # 使用英文的停用词表
tokens = [token for token in tokens if token not in stops] # 去除停用词
tokens = [token.lower() for token in tokens if len(token) >= 3] # 大小写，短词
wnl = WordNetLemmatizer()
tag = nltk.pos_tag(tokens) # 词性
tokens = [wnl.lemmatize(token, pos=get_wordnet_pos(tag[i][1])) for i, token in enumerate(tokens)] # 词性还原
preprocessed_text = ' '.join(tokens)
return preprocessed_text3.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)
# 3、划分数据集
def split_dataset(data, label):
x_train, x_test, y_train, y_test = train_test_split(data, label, test_size=0.2, random_state=0, stratify=label)
return x_train, x_test, y_train, y_test4.文本特征提取
sklearn.feature_extraction.text.CountVectorizer
sklearn.feature_extraction.text.TfidfVectorizer
from sklearn.feature_extraction.text import TfidfVectorizer
tfidf2 = TfidfVectorizer()
观察邮件与向量的关系
向量还原为邮件
# 4、文本特征提取
# 把文本转化为tf-idf的特征矩阵
def tfidf_dataset(x_train,x_test):
tfidf = TfidfVectorizer()
X_train = tfidf.fit_transform(x_train)
X_test = tfidf.transform(x_test)
return X_train, X_test, tfidf
# 向量还原成邮件
def revert_mail(x_train, X_train, model):
s = X_train.toarray()[0]
print("第一封邮件向量表示为：", s)
a = np.flatnonzero(X_train.toarray()[0]) # 非零元素的位置（index）
print("非零元素的位置:", a)
print("向量的非零元素的值：", s[a])
b = model.vocabulary_ # 词汇表
key_list = []
for key, value in b.items():
if value in a:
key_list.append(key) # key非0元素对应的单词
print("向量非零元素对应的单词：", key_list)
print("向量化之前的邮件：", x_train[0])5.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？
# 5、模型选择
def mnb_model(x_train, x_test, y_train, y_test):
mnb = MultinomialNB()
mnb.fit(x_train, y_train)
pre = mnb.predict(x_test)
print("总数：", len(y_test))
print("预测正确数：", (pre == y_test).sum())
print("预测准确率：",sum(pre == y_test) / len(y_test))
return pre6.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
# 6、模型评价
def class_report(pre, y_test):
conf_matrix = confusion_matrix(y_test, pre)
print("=====================================================")
print("混淆矩阵：\n", conf_matrix)
c = classification_report(y_test, pre)
print("分类报告：\n", c)
print("模型准确率：", (conf_matrix[0][0] + conf_matrix[1][1]) / np.sum(conf_matrix))完整代码：
# -*- coding:utf-8 -*-
from sklearn.model_selection import train_test_split
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.naive_bayes import MultinomialNB
from sklearn.metrics import confusion_matrix, classification_report
import numpy as np
import nltk
from nltk.corpus import stopwords
from nltk.stem import WordNetLemmatizer
import csv
def get_wordnet_pos(treebank_tag):# 根据词性，生成还原参数pos
if treebank_tag.startswith('J'): # adj
return nltk.corpus.wordnet.ADJ
elif treebank_tag.startswith('V'): # v
return nltk.corpus.wordnet.VERB
elif treebank_tag.startswith('N'): # n
return nltk.corpus.wordnet.NOUN
elif treebank_tag.startswith('R'): # adv
return nltk.corpus.wordnet.ADV
else:
return nltk.corpus.wordnet.NOUN
# 预处理
def preprocessing(text):
tokens = [word for sent in nltk.sent_tokenize(text) for word in nltk.word_tokenize(sent)] # 分词
stops = stopwords.words('english') # 使用英文的停用词表
tokens = [token for token in tokens if token not in stops] # 停用词
tokens = [token.lower() for token in tokens if len(token) >= 3] # 大小写，短词
lmtzr = WordNetLemmatizer()
tag = nltk.pos_tag(tokens) # 词性
tokens = [lmtzr.lemmatize(token, pos=get_wordnet_pos(tag[i][1])) for i, token in enumerate(tokens)] # 词性还原
preprocessed_text = ' '.join(tokens)
return preprocessed_text
# 读取数据集
def read_dataset():
file_path =r'SMSSpamCollection'
sms = open(file_path, encoding='utf-8')#读取数据
sms_label = [] # 存储标题
sms_data = []#存储数据
csv_reader = csv.reader(sms, delimiter='\t')
for line in csv_reader:
sms_label.append(line[0]) # 提取出标签
sms_data.append(preprocessing(line[1])) # 对每封邮件做预处理
sms.close()
return sms_data, sms_label
# 划分数据集
def split_dataset(data, label):
x_train, x_test, y_train, y_test = train_test_split(data, label, test_size=0.2, random_state=0, stratify=label)
return x_train, x_test, y_train, y_test
# 把原始文本转化为tf-idf的特征矩阵
def tfidf_dataset(x_train,x_test):
tfidf = TfidfVectorizer()
X_train = tfidf.fit_transform(x_train) # X_train用fit_transform生成词汇表
X_test = tfidf.transform(x_test) # X_test要与X_train词汇表相同，因此在X_train进行fit_transform基础上进行transform操作
return X_train, X_test, tfidf
# 向量还原邮件
def revert_mail(x_train, X_train, model):
s = X_train.toarray()[0]
print("第一封邮件向量表示为：", s)
# 该函数输入一个矩阵，返回扁平化后矩阵中非零元素的位置（index）
a = np.flatnonzero(X_train.toarray()[0]) # 非零元素的位置（index）
print("非零元素的位置:", a)
print("向量的非零元素的值：", s[a])
b = model.vocabulary_ # 词汇表
key_list = []
for key, value in b.items():
if value in a:
key_list.append(key) # key非0元素对应的单词
print("向量非零元素对应的单词：", key_list)
print("向量化之前的邮件：", x_train[0])
# 模型选择（根据数据特点选择多项式分布）
def mnb_model(x_train, x_test, y_train, y_test):
mnb = MultinomialNB()
mnb.fit(x_train, y_train)
ypre_mnb = mnb.predict(x_test)
print("总数：", len(y_test))
print("预测正确数：", (ypre_mnb == y_test).sum())
return ypre_mnb
# 模型评价：混淆矩阵，分类报告
def class_report(ypre_mnb, y_test):
conf_matrix = confusion_matrix(y_test, ypre_mnb)
print("混淆矩阵：\n", conf_matrix)
c = classification_report(y_test, ypre_mnb)
print("------------------------------------------")
print("分类报告：\n", c)
print("模型准确率：", (conf_matrix[0][0] + conf_matrix[1][1]) / np.sum(conf_matrix))
if __name__ == '__main__':
sms_data, sms_label = read_dataset() # 读取数据集
x_train, x_test, y_train, y_test = split_dataset(sms_data, sms_label) # 划分数据集
X_train, X_test,tfidf = tfidf_dataset(x_train, x_test) # 把原始文本转化为tf-idf的特征矩阵
revert_mail(x_train, X_train, tfidf) # 向量还原成邮件
y_mnb = mnb_model(X_train, X_test, y_train,y_test) # 模型选择
class_report(y_mnb, y_test) # 模型评价
6.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？
- CountVectorizer：只考虑词汇在文本中出现的频率，属于词袋模型特征。
- TfidfVectorizer： 除了考量某词汇在文本出现的频率，还关注包含这个词汇的所有文本的数量，能够削减高频没有意义的词汇出现带来的影响, 挖掘更有意义的特征。属于Tfidf特征。
- CountVectorizer与TfidfVectorizer相比，对于负类的预测更加准确，而正类的预测则稍逊色。但总体预测正确率也比TfidfVectorizer稍高，相比之下似乎CountVectorizer更适合进行预测。
来源：oschina
链接：https://my.oschina.net/u/4319831/blog/4291592