Abstract
To mine popular accessed Web pages items and find out their association rule from the Web server Log database for junior users providing recommendation service. A novel GEP-based algorithm for mining multiple-layers association rules was presented. Firstly, takes generalizing technology as a way to value fitness function in GEP (Gene Expression Programming). Then, relying on the significant self-search function of GEP, the most optional species was evolved. The frequent items and association rules in the next deeper layers can be mined by using traditional support-confidence method in sub-database. The algorithm improves on the frame of traditional association rule mining and uses a new evolutionary algorithm for mining association rules. Finally, the validity and efficiency of the method are presented by the application in the paper.