FAKE NEWS DETECTION IN MOLDOVA’S INFORMATION SPACE

Abstract

This work is devoted to the development of a system for automatic detection of fake news in Russian, relevant for the region of the Republic of Moldova and its neighboring countries. Initially, a representative dataset of fake and real news was collected from the news sites. Several machine learning models were applied for fake news detection, including Naive Bayes classifier, logistic regression, nearest neighbors, support vector machines, and random forest. The results demonstrated that support vector machines and random forest provide the highest accuracy of classification, reaching 91%, which is an impressive result for such difficult task. The developed system helps protecting society from disinformation, which is especially important in the modern world, where the information warfare and political manipulation have become commonplace.