Research And Design On A Personalized Recommendation System Based On Web Usage Mining

Compares to the huge data on the web, the individual's capacity of reading and grasping is essentially fixed. It's difficult for people to find what they needs on the web, this is the problem of "information blast". In order to resolve this problem, many technologies are put forward to help the user utilize the resource on the web, one of them is web personalization.The personalized recommendation system based on web usage mining has become a hot study point in the field of the web personalization. However, with the rapid development of the web, the dynamic of the web has become a new challenge for web personalized systems. The content of the web sites and the user's browsing interests often change, but the available recommendation algorithms and the personalized system architectures can not effectively solve the problems with the web dynamic. On this issue, the useful exploration and research in the web personalized technology based on web usage mining are processed in this paper.Firstly, web personalized technologies and the development of the personalized technologies based on web usage mining are introduced. And then the features of the data and the process steps in web usage mining are analysed in this paper.Secondly, this paper summarizes the applications of cluster analysis algorithms in personalized systems based on web usage mining. To the challenge of the available clustering algorithms against the web dynamic, this paper models the page accesses information as a undirected graph, proposes a incremental page clustering algorithm based on the undirected graph. Then this paper proposes the definition of page recommendation score and the recommended decision algorithm based on page clusters.Thirdly, the architecture of the typical of web usage mining personalized recommendation systems and the disadvantage of off-line processing in personalized systems are analysed. On this issue, using the recommendation algorithms previously proposed, this paper proposes a new online personalized recommendation system based web usage mining, and introduces the processing stages of the recommendation system.Finally, on the basis of the design for the recommendation system, the recommendation system is implemented as a module of the Apache Web Server, and then the general performance of the personalized system is valuated, including the form of recommendation, the recommendation effectiveness and system response time.