Monitoring Social Networking Platforms to Detect and Filter Fake News using Ensemble Learning

Abstract

Abstract Social networking platforms and news blogs are providing information to the public. Different business, political, and educational communities rely on these news sources for strategic decision-making. It is straightforward to quickly manipulate and spread real digital news to spread misinformation among communities to get a few benefits or relief. Therefore, an automated system is vital that can detect fake news early during monitoring before it is published online. Several studies have been conducted to detect fake news, focusing on resource-rich languages (mostly English). Because of a lack of annotated corpora, resource-poor languages such as Urdu have not been studied. The objective of this study is to provide an effective method for fake news detection from social media platforms in Urdu. Therefore, in this study, we propose a four-level methodology and perform extensive experiments to find out the best model for fake news detection from social media contents in Urdu. This study proposes a public corpus of Urdu news articles and a methodology for detecting early Urdu fake news. We apply eight machine learning and ensemble learning techniques to three Urdu news corpora. Our experiments show that Bagging with Decision Tree as base learner outperforms the others and obtained F-measure scores of 80.9% on UFN, 84.2% on BET, and 86.02% on FNAC.