Evaluation of Machine Learning Methods for Fake News Detection

Abstract

In a cyber-connected world, fake information appears to be more enticing or interesting to the audience because of their limited attention spans and the plethora of content choices. Taking this into account, fake news detection/classification is definitely becoming of paramount importance in order to avoid the so-called reality vertigo, preclude misinformation and protect actual reality. This chapter presents a comprehensive performance evaluation of eight machine learning algorithms who perform fake news detection/classification based on regression, support vector machines, neural networks, decision trees and Bayes theorem. In every case, our study reaffirms that performance is governed by the nature of data, nevertheless, it sheds light and draws safe generic conclusions with respect to the dimensionality that each algorithm should have, the kind of training that should be performed beforehand for each one of them, and finally the method for generating vector representations of textual information. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.