Research Of Buildig User Interesting Model

Data Mining is a technology that obtains valid, novel, potentially useful and ultimately understandable patterns of non-trivial process from databases, data warehouses or other repositories of large amounts of data. It is one of the most dynamic areas which in artificial intelligence and databases. When the data mining technology applied in the network environment, it is called Web Data Mining. Web Data Mining is from Web users to browse the relevant resources and extract behavior and interest, useful patterns and implicit information. The various forms of documentations and user access to information constitutes a Web data mining objects. The various Web content determines the diversity of Web mining tasks. According to the different data mining objects, Web data mining can be divided into Web content mining, Web structure mining and Web usage mining. One is Web usage mining, also known as Web log mining, or Web user access pattern mining, mining object is information which users leave in Web server, establish user interest model through user interest modeling algorithm, mining users` interests from the information, in order to provide users a better browsing experience. Among the many methods of modeling, association rules and Markov models are two important modeling methods.In practical problems, the same project between different data or the different items between different data will exist some links, association rule is the method used to find these links. But as the amount of data increases, the contradiction between the accuracy of data redundancy and the accuracy increased. So mix the methods of inter-transaction association rules and Markov Model to solve these problems.In this paper, firstly propose the intra-transaction association rules and the inter-transaction association rules, as well as the related Markov model algorithm for systematic analysis and summary, and then based on this, propose a new mining algorithm to solve the corresponding problem:Firstly, propose the inter-transaction association rules mining algorithm based on maximum frequent item sets, by improved Mafia algorithm, get the maximum frequent item sets with the corresponding set of total users, named Common User Intersection(CUI), through conversing the maximal frequent item set of intra-transaction to inter-transaction, analyze the relationship between the different users, analyzing user access to different pages on the Web site, directly found that the association rules between the different users to predict the users` interests. The experiment proved that the method can predict a user interested in a more comprehensive page, better to provide pages users may considered about.Secondly, base on mining inter-transaction association rules with maximum frequent item sets, combining with two kinds of methods to establish users` database, proposed a model based on 2nd Markov model and association rules named DUIM.In addition, this paper presents some new methods and improvements of using association rules and Markov Model to establish user interesting model. Propose the concept of the user point by analyzing the relationship between users with similar interests and thus mapped to the corresponding user data, so that the results can better to satisfy the users` need. Propose an improved Mafia algorithm. Makes this maximum frequent item sets algorithm can find the maximum frequent with CUI(as above), thus make possible to make this algorithm on the transaction between the association rules. Combing Markov Model with the method proposed above. Inter-transaction association rules and 2nd Markov model run together to increase in the accuracy of mining results.