• Function：Input the news article URL of Gunosy and return the category classified by Naive Bayes Model.

• Docker 18.09.2
• Python 3.6.8
• Django 1.11

Automaticlly PEP8 check by travi-ci

Collect and train model with 3200 news articles (8 categories × 20 pages × 20 articles) from 8 categories (エンタメ,スポーツ,おもしろ,国内,海外,コラム,IT・科学,グルメ) of Gunosy news web site https://gunosy.com/ in Get articles.
$docker-compose run app python traindata.py

$git clone https://github.com/Jellytrial/news_classification_webapp.git

$cd news_classification_webapp

In oder to build container, make sure you have installed docker.
$docker-compose build

$docker-compose up

After launching, please input following URL in browser:
http://127.0.0.1:8000/
Then input news article URL of Gunosy, it will return the category.

a. Data have to be transformed to the form for evaluation at first in trans_data.
$docker-compose run app python trans_data.py
b. Then model can be evaluated by command: $docker-compose run app python evaluation.py

All classifiers are evaluated with cross validation score in other models.

Following models are evaluated with 3200 data size.

a. Increase data size from 800 to 3200，and accuracy grows from 0.7989 to 0.9434.
b. Increase training data from words of article title to words both in title and article content.
c. Transform news contents to category-word frequency form by words count.