零基础入门NLP赛事 - Task3 基于机器学习的文本分类

方法一：CountVectors + RidgeClassifier

# Count Vectors + RidgeClassifier

import pandas as pd

from sklearn.feature_extraction.text import CountVectorizer
from sklearn.linear_model import RidgeClassifier
from sklearn.metrics import f1_score

train_df = pd.read_csv('../data/train_set.csv', sep='\t', nrows=15000)

vectorizer = CountVectorizer(max_features=3000)
train_test = vectorizer.fit_transform(train_df['text'])

clf = RidgeClassifier()
clf.fit(train_test[:10000], train_df['label'].values[:10000])

val_pred = clf.predict(train_test[10000:])
print(f1_score(train_df['label'].values[10000:], val_pred, average='macro'))
# 0.74

`结果：

方法二：TF-IDF + RidgeClassifier

# TF-IDF +  RidgeClassifier

import pandas as pd

from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.linear_model import RidgeClassifier
from sklearn.metrics import f1_score

train_df = pd.read_csv('../data/train_set.csv', sep='\t', nrows=15000)

tfidf = TfidfVectorizer(ngram_range=(1,3), max_features=3000)
train_test = tfidf.fit_transform(train_df['text'])

clf = RidgeClassifier()
clf.fit(train_test[:10000], train_df['label'].values[:10000])

val_pred = clf.predict(train_test[10000:])
print(f1_score(train_df['label'].values[10000:], val_pred, average='macro'))
# 0.87

结果：

作业：
改变TF-IDF的参数，并验证精度

ngram_range	max_features	max_df	stop_words	f1_score
(1,3)	3000	default	default	0.8721
(1,3)	4000	default	default	0.8753
(1,3)	2000	default	default	0.8603
(1,3)	5000	default	default	0.8850
(1,4)	5000	default	default	0.8849
(1,2)	5000	default	default	0.8864
default	5000	default	default	0.8605
(1,2)	5000	0.8	default	0.8861
(1,2)	5000	1.2	default	0.8864
(1,2)	5000	default	900,3750,648	0.8895

使用其他机器学习模型，完成训练和验证

Logistic Regression(逻辑回归)
(Multinomial) Naive Bayes(多项式朴素贝叶斯)
Linear Support Vector Machine(线性支持向量机)
0.8932985819206637
Random Forest(随机森林)

from sklearn.model_selection import train_test_split
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.feature_extraction.text import TfidfTransformer
from sklearn.naive_bayes import MultinomialNB
from sklearn.svm import LinearSVC

LinearSVC(),
MultinomialNB(),

分类器	f1值
线性SVM	0.8932
朴素贝叶斯	0.6468

# TF-IDF +  RidgeClassifier

import pandas as pd

from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.svm import LinearSVC
from sklearn.metrics import f1_score

train_df = pd.read_csv('data/train_set.csv', sep='\t', nrows=15000)
# 有时单个词语作为特征还不够，能够加入一些词组更好，ngram_range允许词表使用1~3个词语组合
#max_features 限制词表大小 stop_words过滤指定停用词
#max_df 【0.0，1.0】默认值为1.0 当设置为浮点数时，过滤出现在超过max_df/低于min_df比例的句子中的词语
#当设置为正整数时，则是超过max_df的句子
#过滤指定停用词
tfidf = TfidfVectorizer(ngram_range=(1,2), max_features=5000,stop_words=["900","3750","648"])
train_test = tfidf.fit_transform(train_df['text'])

clf = LinearSVC()
clf.fit(train_test[:10000], train_df['label'].values[:10000])

val_pred = clf.predict(train_test[10000:])
print(f1_score(train_df['label'].values[10000:], val_pred, average='macro'))