相关视频地址:
数据地址:链接:https://pan.baidu.com/s/1bWnoL757Sg7JLw8cPNtT9Q 提取码:mbsw
Kaggle IMDB影评者情感褒贬分类问题,Kaggle地址为https://www.kaggle.com/c/word2vec-nlp-tutorial
原文使用的方法是word2vec将词语转为词向量,再用deep learning方式处理,我们这里使用TF-IDF作为特征,用最简单的朴素贝叶斯和逻辑回归尝试
import re #正则表达式
from bs4 import BeautifulSoup #html标签处理
import pandas as pd
def review_to_wordlist(review):
'''
把IMDB的评论转成词序列
'''
# 去掉HTML标签,拿到内容
review_text = BeautifulSoup(review).get_text()
# 用正则表达式取出符合规范的部分
review_text = re.sub("[^a-zA-Z]"," ", review_text)
# 小写化所有的词,并转成词list
words = review_text.lower().split()
# 返回words
return words
# 使用pandas读入训练和测试csv文件
train = pd.read_csv('/Users/Hanxiaoyang/IMDB_sentiment_analysis_data/labeledTrainData.tsv', header=0, delimiter="\t", quoting=3)
test = pd.read_csv('/Users/Hanxiaoyang/IMDB_sentiment_analysis_data/testData.tsv', header=0, delimiter="\t", quoting=3 )
# 取出情感标签,positive/褒 或者 negative/贬
y_train = train['sentiment']
# 将训练和测试数据都转成词list
train_data = []
for i in xrange(0,len(train['review'])):
train_data.append(" ".join(review_to_wordlist(train['review'][i])))
test_data = []
for i in xrange(0,len(test['review'])):
test_data.append(" ".join(review_to_wordlist(test['review'][i])))
train_data
y_train
0 1
1 1
2 0
3 0
4 1
5 1
6 0
7 0
8 0
9 1
10 0
11 1
12 1
13 0
14 0
15 0
16 0
17 0
18 1
19 1
20 1
21 1
22 1
23 0
24 0
25 1
26 0
27 0
28 0
29 0
..
24970 1
24971 1
24972 1
24973 0
24974 1
24975 1
24976 0
24977 1
24978 1
24979 1
24980 1
24981 1
24982 0
24983 0
24984 0
24985 0
24986 1
24987 1
24988 1
24989 1
24990 1
24991 0
24992 0
24993 0
24994 0
24995 0
24996 0
24997 0
24998 0
24999 1
Name: sentiment, dtype: int64
from sklearn.feature_extraction.text import TfidfVectorizer as TFIV
# 初始化TFIV对象,去停用词,加2元语言模型
tfv = TFIV(min_df=3, max_features=None, strip_accents='unicode', analyzer='word',token_pattern=r'\w{1,}', ngram_range=(1, 2), use_idf=1,smooth_idf=1,sublinear_tf=1, stop_words = 'english')
# 合并训练和测试集以便进行TFIDF向量化操作
X_all = train_data + test_data
len_train = len(train_data)
# 这一步有点慢,去喝杯茶刷会儿微博知乎歇会儿...
tfv.fit(X_all)
X_all = tfv.transform(X_all)
# 恢复成训练集和测试集部分
X = X_all[:len_train]
X_test = X_all[len_train:]
# 多项式朴素贝叶斯
from sklearn.naive_bayes import MultinomialNB as MNB
model_NB = MNB()
model_NB.fit(X, y_train) #特征数据直接灌进来
MNB(alpha=1.0, class_prior=None, fit_prior=True)
from sklearn.cross_validation import cross_val_score
import numpy as np
print "多项式贝叶斯分类器20折交叉验证得分: ", np.mean(cross_val_score(model_NB, X, y_train, cv=20, scoring='roc_auc'))
# 多项式贝叶斯分类器20折交叉验证得分: 0.950837239
多项式贝叶斯分类器20折交叉验证得分: 0.950837239
# 折腾一下逻辑回归,恩
from sklearn.linear_model import LogisticRegression as LR
from sklearn.grid_search import GridSearchCV
# 设定grid search的参数
grid_values = {'C':[30]}
# 设定打分为roc_auc
model_LR = GridSearchCV(LR(penalty = 'L2', dual = True, random_state = 0), grid_values, scoring = 'roc_auc', cv = 20)
# 数据灌进来
model_LR.fit(X,y_train)
# 20折交叉验证,开始漫长的等待...
GridSearchCV(cv=20, estimator=LogisticRegression(C=1.0, class_weight=None, dual=True,
fit_intercept=True, intercept_scaling=1, penalty='L2', random_state=0, tol=0.0001),
fit_params={}, iid=True, loss_func=None, n_jobs=1,
param_grid={'C': [30]}, pre_dispatch='2*n_jobs', refit=True,
score_func=None, scoring='roc_auc', verbose=0)
#输出结果
print model_LR.grid_scores_
mean: 0.96459, std: 0.00489, params: {'C': 30}