Web Based Personalized Mining Method

With the rapid development of science and the popularity of Information Superhighway, people have already been involved in the information world. There is a huge mount of information on the internet, the structure of data is isomerous, multiple and distributed. Because the amount of information is increasing very fast, more and more people begin learning from the internet.NERMS(Network Educational Resource Management System)is an important project from Science and Technology department of Jilin Province, which is developed by Jilin University. The aim of NERMS is to efficiently manage and organize the huge amount of Network Educational Resource in order to share and retrieve the resource easily. In NERMS, we provide personalized web pages for each user based on their own profile, changed the web browser from passive receiver to a positive requester, and provided an initiative information service to browsers. This is considered as the key point of the information service of next generation.Personalized initiative information services are implemented by collecting and analyzing the users' requirements, interests and accessing history, construct a profile for users, then apply the filtering and sorting to the users' profile and generate recommendation or push information for users. The Personalized Information Service on the internet must has the following three abilities: users' profile exactly describing the users' interest; users profile dynamically changed with the users' activity; developing new information domain automatically and providing recommendation for users. In this paper, first I introduce the personalized information service recommendation technology. Then describe the GSP algorithm. GSP algorithm is an AprioriAll based algorithm. The introducing of GSP is to discover the patterns within the time constrains, sliding windows. From these appending, we can change the database into an accessing sequence of many users. Each customer sequence lists all the accessing of them. So mining the frequent sequence problem is to find all the successors (itemsets) with enough frequency. Then I introduce the principle and implementation method, and how to apply this method into NERMS. GSP(Global Sequential Pattern)mining method can find the sequential pattern from the web log which stored users' accessing sequence and retrieve the common pattern of the users accessing, gives the user corresponding recommendation. In this paper, I also prove that the advantage of GSP is that the efficiency processing and saving more time than other algorithms. In this system, I have finished all parts of the GSP algorithm, it can give correct output of the frequent sequence pattern of customers efficiently. This algorithm often used to mining transaction information and web log. In this paper I also compare the GSP with the AprioriAll algorithm, we can see that the efficiency of GSP is much better than AprioriAll algorithm.At last, I present the interface and output of the personalized system. Also I list the recommendation time. From these data, we can see that the efficiency of this system is excellent and acceptable.