cjhmpower / 20_newsgroups_classification Goto Github PK

This is a multi-class documents classification task on 20_newsgroups dataset. The packages this project used include scikit-learn-0.18, gensim-0.13.4.1, numpy-1.11.2, matplotlib-2.0.0, nltk-3.2.2,tqdm-4.11.2 scipy-0.18.1 All the code run in Python 2.7. It takes about 90 minutes for the entire notebook to run on an i7-6700HQ CPU. You can skip some cells according to the quotes to save time.

The text8 corpus file was downloaded from http://download.csdn.net/download/z3232186/8753901

The word2vec model has already been trained and saved at file 'word2vec_model'.You do not need to train it again.

The output1.png is the output image after testing all the algorithms performance based on TF-IDF model

The report.pdf is the final project report.

                                                                      ----------------March 20th 2017