A Social Network Based Approach to Detect Fake News on Twitter Data Using Machine Learning

Abstract

The availability of social media together withits enhanced accessibility has led to increasingly fastspread of deceptive information while creating serioustroubles for society and its citizens. The nature of fakenews within modern digital settings creates doubts aboutits effects on public belief and political decisions as wellas democratic functions. Fake news operations alreadyexisted, but technological progress combined with socialmedia platform growth especially among YouTube andFacebook and Twitter users has created ideal conditionsfor fast spreading misinformation. The urgent needexists to investigate how false information spreadsthrough multiple social media platforms because of itsconcerning rate of growth. This study utilizes a socialnetworking detection method that depends on networkproperties through the Communities through DirectedAffiliation (CoDA) algorithm. Different experimentswere performed to validate the proposed approachthrough evaluations on the FakeNewsNet dataset.Experimental findings show an Random forest achivedbest results with accuracy 0.83, F1-score 0.71,precision0.78 and recall 0.64 among all the other modelsfrom the proposed detection methods. Researchfindings from this study expand the field related to fakenews detection through network-based perspectives thatenhance existing methods using content-based andlinguistic analysis approaches.