A Proposed Ensemble Voting Model for Fake News Detection

Fake news or rumors are a phenomenon that significantly influences our social lives. Politicians in the political world usually rely on fake news as a powerful mechanism to change public opinion. Fake news spread through the media poses a real threat to the credibility of information, and the detection of fake news has attracted increased attention in recent years. Therefore, it becomes highly necessary to develop a method to identify fake news. This paper proposes a new ensemble voting model for detecting fake news in online text using a hybrid of machine learning and deep learning algorithms. Our ensemble model consists of three algorithms, namely, Convolution Neural Network (CNN) Gated Recurrent Unit (GRU) model of Recurrent Neural Network (RNN) and Random Forest. We relied on Natural language processing to extract statistical and representative features from the LIAR dataset. We experimented with the extracted features with our ensemble model. Experimental evaluation showed that our model achieves the best performance on the LIAR dataset with an accuracy of 0.410.