[PYTHON] Text filtering with naive bayes in sklearn

from sklearn.datasets import fetch_20newsgroups Let's use the dataset of.

Code to output the category

from pprint import pprint
pprint(list(newsgroups_train.target_names))

This seems to be an article category, but at the beginning it was unclear what was written.

['alt.atheism',
 'comp.graphics',
 'comp.os.ms-windows.misc',
 'comp.sys.ibm.pc.hardware',
 'comp.sys.mac.hardware',
 'comp.windows.x',
 'misc.forsale',
 'rec.autos',
 'rec.motorcycles',
 'rec.sport.baseball',
 'rec.sport.hockey',
 'sci.crypt',
 'sci.electronics',
 'sci.med',
 'sci.space',
 'soc.religion.christian',
 'talk.politics.guns',
 'talk.politics.mideast',
 'talk.politics.misc',
 'talk.religion.misc']

Upon investigation, it turned out to be a net news protocol.

Newsgroup reading fj.comp.applications.excel, fj.comp.oldies, fj.comp.misc, fj.os.ms-windows.win95, fj.os.msdos, fj.net.providers, fj.net.words, fj.life.hometown.hokkaido, fj.jokes.d, fj.rec.autos, fj.rec.motorcycles, fj.news.group.*, fj.news.policy, fj.news.misc, fj.news.adm, fj.news.net-abuse, fj.questions.fj, fj.questions.internet, fj.questions, misc, fj.sci.chem, fj.engr.misc http://www2s.biglobe.ne.jp/~kyashiki/fj/arukikata/WonderfulFj.html

Network News Transfer Protocol was news that used fj (news group).

sample.py

import numpy as np
from sklearn.datasets import fetch_20newsgroups
from sklearn.naive_bayes import MultinomialNB
from sklearn.feature_extraction.text import CountVectorizer
import nltk
from pprint import pprint

def stopwords():
    symbols = ["'", '"', '`', '.', ',', '-', '!', '?', ':', ';', '(', ')', '*', '--', '\\']
    stopwords = nltk.corpus.stopwords.words('english')
    return stopwords + symbols

newsgroups_train = fetch_20newsgroups(subset='train', remove=('headers', 'footers', 'quotes'))
newsgroups_test  = fetch_20newsgroups(subset='test', remove=('headers', 'footers', 'quotes'))

#Show news category
pprint(list(newsgroups_train.target_names))
#Article data
#print(newsgroups_train.data)

#Create CountVectorizer class
vectorizer = CountVectorizer(stop_words=stopwords())
#print(vectorizer)
#Make a vocabulary dictionary
vectorizer.fit(newsgroups_train.data)

# Train
#Substitute document term matrix for X
X = vectorizer.transform(newsgroups_train.data)
# print(newsgroups_train.target)
y = newsgroups_train.target
# print(X.shape)

clf = MultinomialNB()
clf.fit(X, y)
print(clf.score(X,y))

# Test
X_test = vectorizer.transform(newsgroups_test.data) 
y_test = newsgroups_test.target

print(clf.score(X_test, y_test))

Data: Correct answer rate 60% Test data: Correct answer rate 80%

It seems that.

I used it as a reference. It was very helpful. http://qiita.com/kotaroito/items/76a505a88390c5593eba