YisongMiao / CS5228-project

Winning 2nd place🥈at NUS CS5228 in-class Kaggle competition 2018!

CS5228 Final Project - News Article Classification

Project webpage: https://www.kaggle.com/c/cs5228-article-category-prediction

Project instructor:

• Prof. Wang Wei (https://www.comp.nus.edu.sg/~wangwei/)
• PhD Candidate Qi Panpan
• PhD Candidate Zhu Lei

Replicate out experiment

• python model-1.py

  Use this to train model 1 and make prediction. (with only title)

• python model-2.py

  Use this to train model 2 and make prediction. (with title and webpage content)

• python ensemble-rule-based.py

  Our rule based ensemble method

• python check-replication.py

  You can use this to check if you(also we) can replicate our best work on Kaggle, the answer is YES!

Dataset to contribute

We contribute our obtained webpage content, you can use this to push your model's performance to a higher level! Our team is really excited about that!

Usage: It is a pickle file, after load from pickle, it is a list, the index is in the same order as that in the original dataset.

Link:

https://github.com/YisongMiao/CS5228-project/blob/master/data/train_v2.pk

https://github.com/YisongMiao/CS5228-project/blob/master/data/test_v2.pk

Study purpose

• https://github.com/YisongMiao/CS5228-project/blob/master/zjd/first%20version/svm.py

  It is code by Junda to use grid search to find best parameter

• https://github.com/YisongMiao/CS5228-project/blob/master/get-webpage-content/spider.py

  It is code by Chen Song to obtain webpage content

Give us a star on this git repo if you like it!

Contact:

• Junda Zhang: [e-mail] (git: https://github.com/AHEADer)

• Chen Song: [e-mail]

• Yisong Miao: [e-mail]