Fake News Detection Using Time Series and User Features Classification

Abstract

In a scenario where more and more individuals use online social network platforms as an instrument to propagate news without any control, it is necessary to design and implement new methods and techniques that guarantee the veracity of the disseminated news. In this paper, we propose a method to classify true and false news, commonly known as fake news, which exploits time series-based features extracted from the evolution of news, and features from the users involved in the news spreading. Applying our methodology over a real Twitter dataset of precategorized true and false news, we have obtained an accuracy of 84.61% in 10-fold cross-validation, and proved experimentally that all the selected features are relevant for this classification task.