Towards Fake News Identification using Machine Learning

Abstract

The abundance of social media platforms and their usage for news dissemination has gained a lot of attention lately. However, they have some pros and cons. On the one hand, individuals consume latest news instantly and freely, while enjoying the ease of access of instantaneous information transmission. On the other hand, such abundance makes it easy to wide-spread “fake news where the content purposefully incorporates incorrect information to serve some hidden agenda. In this paper, we investigate multiple machine learning algorithms on the road to identify fake news in a proactive manner. We first analyze the viability of applying the Natural Language Processing (NLP) technique to build a labeled dataset. We, then, introduce two approaches for NLP visualization and discuss their performance before selecting the best performer. Using logistic regression, and multinomial Naïve Bayes algorithms, we classify fake news in new data. Finally, we discuss our achieved results and share our lessons learned and recommendations, especially that we achieved an accuracy of 98% in our experiments.