Identifying Fake News in the Russian-Ukrainian Conflict Using Machine Learning

Abstract

The widespread dissemination of false information made possible by social networks' universal accessibility and ease of use may be harmful to both individuals and societies. Regrettably, one of the most popular subjects on social media right now is the crisis between Russia and Ukraine. Spreading fake news about this conflict may cause serious consequences to both countries and their citizens. Therefore, we are motivated to build a fake news detection system similar to those systems already accessible in other fields like healthcare. In this study, we build a dataset of fake and real tweets about the Russian-Ukrainian conflict and evaluate the power of machine learning in this context. The pre-train BERT and five classical machine learning algorithms; namely support vector machine (SVM), Decision Tree (DT), K-nearest neighbor (KNN), logistic regression (LR), and Naïve Bayes (NB) are trained and evaluated through different scenarios of the dataset. The results show that it is possible to develop a system that can discern between real and fake news regarding the Russian-Ukrainian conflict. Support vector machine and logistic regression outperform other learning algorithms and produce comparable prediction models with approximately 76% of accuracy.