Research On Web Prediction Model Based On Web-Log

With the remarkable and exponential growth rate of Web information and users, how to reduce the user perceived access latency and improve the quality of service of the network is coming a crucial problem, and Web prefetching and Web caching are the primary solutions. Web caching technique has been widely used in different places of Internet. But as dynamic documents and personal services increase all over the world, the performance of caching deteriorates significantly. As a result, Web perfecting, which is an efficient way of making up for Web caching, and the most effective method to break the upper bound of caching performance----is coming a hotspot in Web speedup research area.The Markov model is a simple and practical tool to prefetch Web. But some existing prediction methods based on Markov model still have some shortcoming. So it becomes a new lesson in the area of Web log mining that how to improve prediction methods. This paper analyses the current domestic and international research results of how to use Markov model to predict Web. Then we find some problems of existing prediction methods based on Markov models and we study the improving of prediction methods based on Markov model.First of all, this thesis introduces the development and the state of the Internet and WWW, gives the problems Internet faced and corresponding solutions; and describes the concept, classification of Web data mining; and Web log mining data preprocessing process. In order to overcome the drawbacks of Apriori algorithm for mining frequent itemsets, TIMV algorithm was proposed.Second of all, the interest is the selectivity attitude of objective matter of a person, and measuring user's browse interest exactly is the base of Web base of Web schema mining. This paper analyses the present the shortage of the style of measure and expresses the browsing interest of user. For instance, the too simple measure fashion often leads to difficulty of distribution which is the user interested in or not, not considering the page information amount's influence on the users' browse time and so on. As a result, point out a method based on users' browse behavior to measure the users' browse interest.Then, a hybrid Markov predictor model was put forward based on the step-2 Markov model, which can solutes the problem of high memory demand and the low applicability. Besides that, this paper gives the sustaining theory and the way to get the parameters.Finally, experiments have been made based on the prediction model and experimental results are analyzed.