Rumour detection on social media through Ensemble deep learning model

Abstract

The widespread dissemination of several rumours and fake news has extremely affected many business organizations. Rumours or fake news spread and go viral on social websites causing significant social, business and political problem around the world. Fake news distorts public belief, affects companies reputation, disturbs community, reduces government integrity, makes business vulnerable and creates an enormous danger to the social constancy. This paper proposes a novel algorithm to detect rumour spread through tweeter. Various advanced machine learning techniques: Random Forest, Gradient Boosting, Adaboost and Hard Voting classifier are applied to detect the rumours from online social media through tweeter and Facebook. The proposed model is based upon the CNN-BiLSTM-BiGRU approach which efficiently classifies tweets among rumours and non-rumours category. The proposed model provides 99% accuracy and outperforms other baseline models like Random Forest, AdaBoost, Gradient Boosting, and Hard Voting. This hybrid recommender framework can be a more beneficial system to detect rumours at earlier stages from online platforms.