Modeling disseminating behavior based on explicit duration recurrent networks for microblog misinformation detection

Microblog users are consumers of information, and also disseminators. This study presents a new misinformation detection method based on disseminating behavior. The explicit duration recurrent network (EDRN) is adopted to construct the disseminating behavior model. The EDRN is a recurrent neural network based on an extended hidden semi-Markov model. The observation is based on the opinion and confidence level of the disseminator. The disseminator’s confidence level is estimated by the support vector machine, and the disseminator’s opinion is inferred by sentiment analysis. Experiments are conducted based on real datasets of Weibo. The experimental results indicate that EDRN is superior to popular recurrent neural networks in terms of accuracy, F1, and recall.