Demystifying black-box learning models of rumor detection from social media posts

Abstract

Social media and its users are vulnerable to the spread of rumors, therefore, protecting users from the spread of rumors is extremely important. For this reason, we propose a novel approach for rumor detection in social media that consists of multiple robust models: XGBoost Classifier, Support Vector Machine, Random Forest Classifier, Extra Tree Classifier, Decision Tree Classifier, a hybrid model, deep learning models-LSTM and BERT. For evaluation, two datasets are used. These artificial intelligence algorithms are often referred to as "Blackbox"where data go in the box and predictions come out of the box but what is happening inside the box frequently remains cloudy. Although, there have been several works on detecting fake news, the number of works regarding rumor detection is still limited and the models used in the existing works do not explain their decision-making process. We take models with higher accuracy to illustrate which feature of the data contributes the most for a post to have been predicted as a rumor or a non-rumor by the models to explain the opaque process happening inside the black-box models. Our hybrid model achieves an accuracy of 93.22% and 82.49%, while LSTM provides 99.81%, 98.41% and BERT provides 99.62%, 94.80% accuracy scores on the COVID19 Fake News and the concatenation of Twitter15 and Twitter16 datasets respectively.