Exploring News-Feed Credibility using Emerging Machine Learning and Deep Learning Models

Abstract

In the recent past, the phenomenal growth, availability, and access of information on social media have made it perplexing to discern between real and fake information. The faster and easier dissemination of information through various means has accelerated the explosive growth of its falsehood. At the same time, the newsfeeds and their credibility in the social networks are in danger since the fake news is alarmingly disseminating very fast. Henceforth, the credibility of the newsfeeds or any information has become a real research challenge to cross-check the respective news or information. The cross verification can be performed concerning its source, the exact content, and the respective publisher to catalog it into fact or fake. Despite a few constraints, machine learning plays a crucial and significant role in classifying the respective news feeds. Various machine learning methodologies such as BLR (Bilinear logistic regression), NB (Naive Bayes), SVM (Support Vector Machine), and RF (Random Forest) have been reviewed and experimented with for detecting the fact and fake news feeds. After the experimentation, the limitations of the respective machine learning methods were explored and noted. Henceforth, a deep learning method called BERT is implemented and it is observed that BERT provides better efficiency than the machine learning algorithms. Furthermore, it is ascertained that the deep learning method provided the best accuracy of 99.79 % with the available dataset.