A Personalized Recommendation System Based On Multi-level Association Rule In E-commerce

As the Internet era gradually develops to be the Internet+big data era,information resources become very richer. How to obtain the needed information from mass resources accurately and effectively needs to be addressed.Personalized recommendation is an effective method to solve the problem of information overload.This dissertation is mainly focused on multi-level association rules mining and the user interest model based on data mining. The aim is to provide better personalized recommendation service and design a personalized recommendation system based on E-commerce. The main research contents are as follows:(1) An improved algorithm CTE-MARM(Constraint Transaction Extension-Multi-level Association Rule Mining) based on FP_Growth algorithm is presented, which is to solve the problems of traditional multi-level association rule mining, i.e., inefficient execution, high redundant rules. K-levels constraint extension improves the execution efficiency and reduces the redundancy on the compressed transaction sets along with multi-level association rule mining,where it applies to the different user needs and application scenarios.(2) Association rules are introduced into the user interest model. With combination of the rules generated by multi-level association rules mining and user's interests generated by the study of user's browsing behavior, user interest model is built based on the E-commerce sites. This proposed model is able to provide better services according to user's interest and shopping trends of all the consumers.(3) A personalized recommendation system is put forward. The user interest model is drawn based on association rule into the system. Experimental results show that the designed recommendation system has much better services by combination of user's interest and shopping trends of all the consumers.