Identifying Fake News written on Albanian language in social media using Naïve Bayes, SVM, Logistic Regression, Decision Tree and Random Forest algorithms

Abstract

Undoubtedly, one of the things more valuable nowadays is information. But, the information on which form we give it has a very important role for a society, where this has a wide impact especially some topics at the time when the flow of information could be unaffordable. Therefore, not every piece of information we take should be taken as true, in the time we live in, it has a lot of sources where share lots of fake news, and many times they get out of control. Hence, this study aims to identify which algorithms classify in better form these fake news from different sources. By using semiautomatic techniques of extracting data from the electronic source we prepared 100 articles/ titles and we used them to analyze five algorithms such as Logistic Regression (LR), Naïve Bayes, Support Vector Machine (SVM), Random Forest and Decision Tree. According to state-of-the-art, these algorithms are the most used in this field by classifying, the fake news and the better accuracy. In our experiments, we have used Albanian sentiment and the best accuracy has shown Naïve Bayes with 66% then it is Decision Tree with 60% and SVM with 53%. All of the results we have compared to each other and we presented the results.