Disinformation detection using passive aggressive algorithms

Abstract

Disinformation, also as known as fake news, is overwhelming. Intentionally false information is widespread. However, the detection of fake news is remaining to be a challenge due to the nature of the complexity of languages. Linear regression algorithms are proven to be effective in many practices. In this paper, the Passive-Aggressive and the Multinomial Naive Bayes are studied for fake news detection that involves term frequency and inverse document frequency to vectorize news content. Our results show that the Passive-Aggressive is more efficient than Multinomial Naive Bayes by combing with term frequency and inverse document frequency and could be applied as a primary screen for complex disinformation detection practically.