Deep Learning Approaches for Detecting Rumors on Social Platforms: A new Detection Model

Abstract

The rapid expansion of social platforms has led to an unprecedented flow of user-generated content, where rumors can spread rapidly and cause significant social, political, and economic harm. Efficient detection of such misinformation has therefore become a critical challenge. In this study, we first conduct a comprehensive systematic review of existing rumor detection techniques, covering both traditional machine learning and advanced deep learning approaches, and highlight key trends and limitations. Building on these insights, we formulate rumor detection as a text classification problem and propose deep learning–based solutions. We evaluate several architectures, including convolutional neural networks (CNNs), recurrent neural networks (RNNs), long short-term memory (LSTM) networks, and transformer-based models. Experiments on real-world rumor datasets demonstrate that deep learning methods consistently outperform conventional approaches, achieving higher accuracy, robustness, and generalization. These findings confirm the efficiency of deep learning for rumor detection and underscore its potential for deployment in scalable social media monitoring systems.