Dissertation
Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer software > Program design,software engineering > Programming > Database theory and systems > Database systems : by type

Reserach on Personalization Technologies Based on Web Usage Mining

Author SongXiaoHui
Tutor ZhangZhongPing
School Yanshan University
Course Computer Software and Theory
Keywords Web usage mining Personalization technology Effective access sequence Maximum frequent set Personalized recommendation
CLC TP311.132
Type Master's thesis
Year 2009
Downloads 96
Quotes 0
Download Dissertation

With network resources enriching continuously and network information increasing rapidly, information overloadx and resource lost are becoming the bottleneck to the people using Web information effectively. A new service model called Personalized Service, which can automatically organize and arrange the information according to users’ interest, need to be introduced by the Website. It makes information service from one-to-multiply in tradition to one-to-one, and using web usage mining in the personalized service has become a research focus. The internal and external researches of research status on personalized technology are analyzed and compared, and the problem of personalized service based on web usage mining is researched from a new point of view.Firstly, the actual access sequence (effective access sequence) is introduced in this paper, which allows pages can be clicked on repeatedly and makes the pages in the itemset adjacent or equipotential adjacent, and then a mining algorithm of maximum frequent itemsets is put forward. This algorithm adopts the methods of overlapping, uniting and sifting for mining maximal frequent itemsets, and also the detection strategies of majorized subset and neglected single page in order to improve its execution performance.Firstly, the actual access sequence (effective access sequence) is introduced in this paper: repeatability, which allow pages can be clicked on repeatedly; continuity, which makes the pages in the itemset adjacent or equipotential adjacent; concomitant, which mine maximum frequent itemset in the set, and then a mining algorithm of maximum frequent itemsets is put forward. This algorithm adopts the methods of overlapping, uniting and sifting for mining maximal frequent itemsets, and also the detection strategies of majorized subset and neglected single page in order to improve its execution performance.Secondly, a new directed graph structure of maximal frequent sequential pattern and a personalized recommendation algorithm based on it is raised. As only access some subgraph in the directed graph without searching the whole pattern database in order to shorten pattern-matching time greatly, so the method can satisfy real-time page recommendation.Finally, the feasibility and validity of maximum frequent itemset mining algorithm and its application in the personalized recommendation is checked by experiment, and the efficiencies of the two methods are compared respectively.

Related Dissertations
More Dissertations