An Adaptive Prediction Model Of Online Topic Diffusion Trend

The rapid development of Internet and recent explosion of online media such as blogs and Twitter facilitate the fast diffusion of online topic.These topics which diffuse fast can attract attention of most people and they usually have an impact on social stability.Therefore,how to predict online topic diffusion trend becomes a more and more important problem.The process of topic diffusion is the result of many factors especially some uncertain factors.Hence,it is quite a challenging task to accurately predict the trend of online topic diffusion.Based on the analysis of online topic diffusion trend and its characteristics,a one step adaptive prediction model of online topic diffusion trend is proposed in this paper.This one step adaptive prediction model consists of two parts: chaos prediction and adaptive adjustment.A modified traditional chaos prediction method and an adaptive adjustment strategy are combined to predict the online topic diffusion trend.The results of numerical simulation reveal that this model has a good performance on prediction of online topic diffusion trend.Moreover,this model has many advantages,including small amounts of training data,few parameters,general applicability,and above all higher fitting accuracy compared with other methods.In addition,a multi step adaptive prediction model of online topic diffusion trend is proposed too.This multi step adaptive prediction model is based on the one step adaptive prediction model.Combined with the prediction results of one step adaptive prediction model and a multi step iteration strategy,this multi step adaptive prediction model can predict the future several steps of online topic diffusion trend preferably.In the experiment,the prediction results of numerical simulation show a good agreement with the real data.