Abstract: According to the novel aspect of natural hierarchical property of Web site， the inverted index structure was proposed based on Hash table (IIS-HT) to promote the speed of data preprocessing. Based on IIS-HT， a prediction model was also proposed which was based on statistics to predict users’ browsing behavior. The design idea， definition， framework and key construction methods of the model were also given. Finally， the proposed model was tested with real data. The experimental results show that the model and prediction algorithm could reduce the scope of candidate pages and improve the speed of prediction with adequate accuracy.

Key words: Web site, users’ browsing behavior prediction, Markov model, Bayesian theorem