Content-based fake news classification through modified voting ensemble

Credibility is a crucial element for journalism. As fake news impacts credibility, it affects the general public, policymakers, decision-makers, and the journalistic environment. However, current research on fake news using content-based approaches focuses majorly on one or two dimensions of stylometrics, semantic and linguistic processes, but not on these three simultaneously. Considering that content-based detection of fake news would benefit from a multidimensional approach because of their inherent characteristics, we proposed a method that uses all of these dimensions to improve classification accuracy, using a voting ensemble designed in an ensemble classifier form. The results show that the multidimensional voting classifier has produced more accurate results than its peers while being more sensitive to distinguish between true and false news when using randomized data.