Research On Web User Access Pattern Online Mining Recommending System

Mining the pattern of visiting websites is one of the most important applications in present data mining. The user access pattern mining can make the websites builders understand clearly different interests of their users and the visiting situation of their whole web pages. On the other hand, in order to facilitate different types of user accesses, it can adjust endlessly the logic organizing structure of the web or set up self-adaptive website. At present, this kind of mining technology includes two category , one is web usage mining. It does analytical mining for the Web behaviors of the Internet users to obtain the inherent system describing user access pattern. In this way, it offers the decision-making support for the websites builders to improve their web structure and web content. The other is collaborative filtering technology which can accomplish individual services for the websites. All present user access pattern mining tools have some shortcomings, such as slow mining speed, unsuitable anonymous users, complicated usage, inefficiency, low mobility, much mining limitation and so on.To be aimed at this circumstance, we have integrated present user access pattern mining skill comprehensively and put forward the design of User Access Pattern Online Mining Recommending system. UAPOMR is a online-mining recommended system used on the server terminal of the Website which helps to build the self-adaptive website , and it can serve anonymous users directly. First, UAPOMR system forms user profiles according to the historical visiting pattern. Then it implements efficient recommended algorithm by comparing users' active visiting page sequence and users' profiles to produce a recommended page which to offer to the user.The recommended algorithm of UAPOMR system includes recommendation based on transaction_clusters and recommendation based on association rules clusters. The former one gains the aim of improving the recommended speed by grouping the transaction clusters, which make system has ability to solve the limitations of the collaborative filtering technology which can't process huge data. The latter one adopts hypergraph partitioning skill and effectively groups the association rules according to the users' historical visiting data. And it can find the same visiting pattern of the different interests-oriented user groups to enhance the quality recommended, by means of adjusting the number of clusters , UAPOMR system solve the contradiction between mining speed and nicety effectively for recommending in real-time.In our UAPOMR system, we consider time feature, web sort feature, Web pages' physical hyperlink distance ,and it make our recommended algorithm more efficiency.We introduce the whole design of UAPOMR system comprehensively and in detail, and evaluate the recommended algorithm by experiments.