Improving Prediction of Arabic Fake News Using Fuzzy Logic and Modified Random Forest Model

Abstract

Throughout the last few years, the world is witnessing the so-called age of social media, as there is a complete dependence on these sites for following up on events and activities. The problem is that the misinformation or fake news is always released at the appropriate time, so this false news spreads quickly and takes a very wide resonance. Although several studies are performed to determine English fake news, the identification of Arabic misinformation remains underdeveloped. This study aims to build an improved learning model for detecting fake news in the Arabic language. Unlike previous studies that depended on analyzing the content of the tweet only, this study focuses on the text, user features, and text features. Regarding the content of the tweet, the TF-IDF method was used to convert the words into features and then determine the features that have a high rank. In contrast, a fuzzy model was used to determine the relevant features for the user. Finally, the random forest algorithm has been adapted and improved, and its results are better as compared to other machine learning methods. The accuracy of Improved Random Forest is (0.895), whereas the accuracy of Naive Bayesian and SVM techniques are found to be (0.809) and (0.848), respectively.