Personalized Recommendation Algorithms Based On The Fuzzy Interest Pattern

Electronic commerce has born half a century. The second tide of network-economy is rising and falling.Through this years,this viewpoint has been accept that commerce is the center of all actions ,and the client's serving is the key of successed commerce.serving a growing client demands is the core task for electronic commerce.To be Personalized, how to personalize is key of competition.The personalized recommendate technology is your wanted.In the current,based collaborative filtering(CF) and based on content technology are the major .And the item-based and user-based(User-to-User) filtering are mainstream of CF technology.The tradition user-based filtering hasn't explicit user-model,but compare the users' similar directly to recommend,the result of recommand is not good.Item-based recommanded system overcome the part shortcomings of user-based filtering,but it abandon the user's ontology model ,and may be cause to same recommanded.With these facts,this paper directed mainly the essence of personalize ,and meld the anknowlege of fuzzy mathematics to rise a kind of new model based on the fuzzy interest pattern . At the first,a new way be rised to resolve the question which fuzzy the attributes of commodity. then ,though mining the lists which users selected,we create a single-kind multi-levels fuzzy user model. The second,considering the needed for multi-commodity recommend,we discusses how to build a new multi-kinds user model which include new interest points.Direct the null-sets which user unselected, this paper use the algorithms of mining fuzzy association rules to attempt to get the new interest points. At the last ,a new multi-kinds fuzzy interest model (MFIM) include new interest points has been creat.This is the major contribution in this paper.Because the different between tradition recommended and MFIM-based recommended,a new algorithm is creat for MFIM-based recommended which can find new interest points.Finally,it shows validity in the experimentation,and has effectiver than tradition recommended. To instance this algorithm,multi-agents are bring into experimentative system.The cooperation among of agents have high efficiency to complete recommmend active and to improve creating a society system finally.