Improving Generalizability of Fake News Detection Methods using Propensity Score Matching

Abstract- Recently, due to the booming influence of online social networks, detecting fake news is drawing significant at- tention from both academic communities and general public. In this paper, we consider the existence of con- founding variables in the features of fake news and use Propensity Score Matching (PSM) to select generaliz- able features in order to reduce the effects of the con- founding variables. Experimental results show that the generalizability of fake news method is significantly better by using PSM than using raw frequency to se- lect features. We investigate multiple types of fake news methods (classifiers) such as logistic regression, random forests, and support vector machines. We have consis- tent observations of performance improvement.