1Dept. of Computer Engineering, Liyan Institute of Education, Bushehr, Iran
2Dept. of Computer Engineering, Firoozabad Branch, Islamic Azad University, Firoozabad, Iran
3Dept. of Computer Engineering, Liyan Institute of Education, Bushehr, Iran
Rumors spread on social networks can sometimes have serious negative social effects, and it is usually impossible for humans to manually check the millions of posts created. Therefore, an automated technique for detecting rumors on social networks has a high practical value. In this paper, the normal behaviors of users are analyzed through their posts in the Sina Weibo social network dataset to identify rumors. We offer a autoencoder-based model with hidden multi-layer configuration to automatically detect rumors. The number of different hidden layers have been investigated to evaluate the performance of the model. The output of the autoencoder model is analyzed based on the SIR model for more accurate detection of rumors. Then, according to the rumor spread rate of the SIR model, the rumors are controlled by the SIHR model based on the steady state of the system. Analysis of the SIHR rumor spread model for a social network is based on the degree of each node, where a new node state transfer function is designed for this purpose. The results show better performance of the proposed method with 95.7% detection accuracy than DLSTM and DGRU methods on the Sina Weibo social network dataset.
Social Networks, Rumor Detection, SIR Model, SIHR Model, Autoencoder Model
