Low state-space complexity and high coverage Markov browsing forecast

Abstract

Browsing the World Wide Web (WWW) involves traversing hyperlink connections among documents. The ability to forecast browsing patterns can solve many problems that face producers and consumers of WWW content. Although Markov models have been found well suited to forecasting browsing modes, they have some drawbacks. To solve them, we present a new model, Markov tree model (MTM), to forecast user-browsing modes. It aggregates user-browsing information by a tree. By this structure, a forecast model can't generate an explosive number of states. All the forecast process can be performed on the MTM. During the forecast procedure, a recursive process is adopted to handle the problem of low coverage. If a higher sequence can't get a result, a lower sequence may be used. Experiments confirm that MTM can get higher coverage and lower state complexity. It can be widely used in prefetching, link prediction and recommendation, etc.