Reasearch On The Web Personalization On Makrov Model

With the fast development of Internet, the number of Web pages and the Internetinformation is increasing everyday. More and more Web users get lost in the millionsof the Web information. People have imminence needs to find out the “interested”information from the user browsing log. By mining Web usage information efficiently,we can obtain the knowledge about user access WWW, which can be used for Webservice providers and Web users. The results can provide kinds of information forimproving Web server’s design, improving website server performance, improvingpersonalization service, and so on. This is the Web personalization technologyresearch content and the purpose of the study it.Now, Web personalization has become a new and important research filed in theworld. Its research has important practical significance. And the establishment of aneffective Web user browsing behavior model is the key technologies of study of Webpersonalization. Markov chain is a classical model of Web user navigation predicting.Firstly, this paper summarizes and analyses the characteristics of Web usage miningbased on Markov model. And then the current research status in this field is alsoclassified and introduced. Major contributions are listed as follows:1. A new hybrid Markov navigation prediction model is proposed.In HPG model, though Sate Cloning improves the Accuracy of the higher ordernavigation prediction, only consider the statistics information of users’ log. While theNG model considers the Web pages relative importance using PageRank algorithms,but the precision of higher order is lower. In this paper, we propose a new thought tocombine the two advantages of the both models.2. Propose a Flow Markov Model for approximately compute multi-orderMarkov Chain in navigation predictingIn this paper, we propose a new model for approximately compute multi-orderMarkov chain based on two new personalization principles, which is using Flowmodel to compute higher order Markov chain model based on the first order Markovchain (FMM).The experimental results show that FMM is efficient and effective and can provide objective pages to user. In particular, it has lower space complexity thantraditional Model.3. Improve on Flow Markov ModelFlow Markov Model is a little simple because it just considers the statisticsinformation of Web log and not analysis the structure of Website. So we proposed aimproved model for FMM (IFMM). IFMM computes the relative importance of theweb pages using Pagerank algorithms and analyses the unique user’s navigationbehavior. The experimental results show that IFMM is efficient and effective and canprovide objective pages for users than traditional Model and FMM.Finally, this paper summarized the author’ works and discussed the future works...