Fake News Detection Using Machine Learning

6 Pages Posted: 3 Oct 2019

See all articles by Lilapati Waikhom

Lilapati Waikhom

National Institute of Technology, Arunachal Pradesh

Rajat Subhra Goswami

National Institute of Technology, Arunachal Pradesh

Date Written: October 2, 2019

Abstract

Nowadays most of the people prefer the internet to access news as it is easy and cheap, but that results in wide spreading of fake news very fast. Fake news is often written with an ulterior motive to gain financially, politically, etc. with most of the time having a catchy headline which attracts users or it may also be accidental. But it affects so much to the people. Fake news detection has become a challenging topic nowadays. In this work, we use the LIAR dataset which is collected from POLITIFACT.COM for fake news detection and it is publicly available for use, which provide links to source documents for each case. In all the previous works, the accuracies are all around 30 percent on this dataset. In this work, we use model ensemble techniques to have better accuracy in predicting fake news using the LIAR dataset. We have also tried to simplify the problem statement into binary classification and deployed the same ensemble techniques to have an even better realistic approach for accurate calculation.

Keywords: Ensemble, Fake News, Liar dataset, Classification, XGBoost

Suggested Citation

Waikhom, Lilapati and Goswami, Rajat Subhra, Fake News Detection Using Machine Learning (October 2, 2019). Proceedings of International Conference on Advancements in Computing & Management (ICACM) 2019, Available at SSRN: https://ssrn.com/abstract=3462938 or http://dx.doi.org/10.2139/ssrn.3462938

Lilapati Waikhom (Contact Author)

National Institute of Technology, Arunachal Pradesh ( email )

Arunachal Pradesh
791112
India

Rajat Subhra Goswami

National Institute of Technology, Arunachal Pradesh ( email )

Arunachal Pradesh
791112
India

Do you have negative results from your research you’d like to share?

Paper statistics

Downloads
1,175
Abstract Views
2,907
Rank
33,567
PlumX Metrics