Internet Users' Psychosocial Attention Prediction: Web Hot Topic Prediction Based on Adaptive AR Model

Web hot topic prediction is now one of the most significant research focus in web data mining, which can reflect the internet users' psychosocial predilection, may greatly benefit us. Markov and neural network are such two typical traditional prediction model, however, the Markov method can neither capture nor express the statistical property of the real data while the computation of neural network is quite complex. In this paper, a new method based on adaptive Auto Regession (AR) model is proposed, the parameter estimation algorithm of this model is referred to as Recursive Weighted Least Square (RWLS) and therefore defines the topic trend according to the model, and the computation is simple and quick. Also included are the advantages and shortcomings of this method.