A framework for fake news detection based on the wisdom of crowds and the ensemble learning model

Abstract

Nowadays, the rapid development of social networks has led to the proliferation of social news. However, the spreading of fake news is a critical issue. Fake news is news written to intentionally misinform or deceive readers. News on social networks is short and lacks context. This makes it difficult for detecting fake news based on shared content. In this paper, we propose an ensemble classification model to detect fake news based on exploiting the wisdom of crowds. The social interactions and the user?s credibility are mined to automatically detect fake news on Twitter without considering news content. The proposed method extracts the features from a Twitter dataset and then a voting ensemble classifier comprising three classifiers namely, Support Vector Machine (SVM), Naive Bayes, and Softmax is used to classify news into two categories which are fake and real news. The experiments on real datasets achieved the highest F1 score of 78.8% which was better than the baseline by 6.8%. The proposed method significantly improved the accuracy of fake news detection in comparison to other methods.