A Robust Rumor Detection on Twitter Using an Enhanced LSTM-RNN Model

Abstract

Social networks have significantly impacted people’s lifestyles, providing a crucial platform for communication and information sharing that can engage millions of users within seconds. While social media offers positive influences on lives and relationships, it also fosters the spread of rumors, resulting in adverse effects such as loss of reputation, credibility, and health. This paper introduces a novel algorithm based on long short-term memory and recurrent neural networks (LSTM-RNN) for detecting rumors on Twitter. The proposed algorithm’s performance is evaluated through extensive simulations using the AUG-RNR-DATA dataset, which includes both rumor and non-rumor tweets from six distinct events. The simulation results demonstrate that our algorithm surpasses the baseline methods, achieving a 0.95 accuracy in distinguishing between rumors and non-rumors on the AUG-RNR-DATA dataset.