Research On Information Diffusion Based On Chaos Theory

Since the social demand of human developed and internet pushed forward rapidly,the social gradually become a platform where people can share and communicate with the others,and more and more hotspots ferment and outbreak in social network platform.How to predict the tendency of network sentiment accurately plays a critical and practical role in supervising and guiding the diffusion of network public event,commodity marketing and flow control.In generally,the traditional research methods are to start with inner social networks,some researchers predict the hot events trends by using epidemic model,influence diffusion model and so on.This article will predict the trends from the macroscopic point of view,and introduce chaotic time series prediction to the prediction of public opinion.The main work of this paper is as follows:1.This paper try to introduce chaotic time series prediction to the prediction of hotspots and analyzed the chaotic characteristics of hotspots in social network from different angles.As the nonlinear system is improving continuously,there is more and more evidence to show that maybe social network has chaotic property.After collecting public opinion data,this paper calculate the time delay parameters and the embedding dimension parameters by using C-C algorithm.The Lyapunov exponent is proved to be positive after calculating small-data method,which proves that the time sequence of hotspots in social network has chaotic features.What's more,correlation dimension can also prove the chaotic characteristic.In the end,the phase space of time sequence of hotspots in media and social network is reconstructed to demonstrate the evolution process of the system in a higher dimension.2.In order to improve the accuracy of the short-term public opinion prediction,a model conbines AOLMM prediction model and RBF neural network is designed,and the method extend univariate time series to multivariate.Regarding to the low prediction accuracy of single-model,after improving the clustering method in RBF neural network,this article combines the advantages of AOLMM prediction model and RBF neural network.The experiments results show that compared with the way of prediction by the unit chaotic time series,multivariate chaotic time series have better predictive effects,which provides a new way for predicting the trends of public opinion.