Feature Analysis and Influence Prediction of Rumor Propagation Based on mRMR and Neural Networks

Online rumors often spread very fast and widely. Many people want to control rumors, but it is very difficult, as we do not know the characteristic of rumor propagation. In this paper, an approach of feature analysis for rumor propagation is presented based on some investigations about the factors influencing the spread of online rumors. Taking the rumors of the COVID-19 as examples, the proposed approach based on mRMR feature selection and BP neural networks filters the factors that may influence rumor spreading, and constructs a rumor spreading model for influence prediction based on the influencing factors. Experimental results show that the model can predict the potential influence of rumors at the early stage of rumor propagation, help to detect rumors with high potential influence, and improve the efficiency of rumor control.