Comparative Study of Machine Learning Algorithms and Text Vectorization Methods for Fake News Detection

The detection of fake news is a crucial task in today's society, given the widespread use of social media and online platforms. In this study, we investigate the application of Machine Learning (ML) algorithms for the detection of fake news. We consider two different datasets of categorized news articles of various sizes and apply various ML algorithms, along with two methods of text vectorization. Specifically, we examine Bag of Words and Tf-Idf, with the use of stemming and with different n-gram values. The resulting vectors are processed by Naive Bayes algorithms, Linear algorithms, Support Vector Machines, and Random Forest classifiers. F1-score and computational time for each algorithm-vectorization combination were recorded. Our results have shown that Linear Algorithms and Support Vector Machines combined with Tf-Idf vectors and n-gram value of (1,2) produced the highest accuracies, with an F1-score up to 96.8%.