Comparative analysis of the main machine learning algorithms for the automatic recognition of fake news

Abstract

In the digital era, change is everywhere. The benefits the internet brings are embraced, but, at the same time, challenges brought by technological evolution are needed to be dealt with. One of them is disinformation and the speed of spreading fake news. Whatever the goal, artificial intelligence, machine learning, deep fakes, and voice biometrics are powerful tools to develop fake news, and threat actors use them often. Therefore, machine learning has become a countermeasure, an instrument to combat the fake news phenomenon. This research examines a number of machine learning algorithms to determine which is the most accurate for automatically recognizing fake news from the Politics domain. The results show that TF-IDF can be used in preprocessing the dataset, and the Passive Aggressive SVC and Random Forest algorithms show the best performance for a given fake news dataset.

Keywords

• ai
• machine learning
• fake news
• disinformation
• algorithms testing
• accuracy