E-commerce Personalized Recommendation System Research

Today, in the information age, as the popularization of Internet and the quick development of information techonology, the E-commerce system has become more and more mature, and it has given a lot more service to all the users, but its structure has become more and more complicated. Users might get lost in the sea of commodity information, and couldn't find those things they are in need of. Personalized recommendation system for E-commerce has then born in the right time. It can search through the E-commerce platform according to user's information and behavior, and then recommend the target products to help finishing the purchase procedure.The system can collect enormous user data, such as user trading data, user registration data, user evaluation data and user voting data etc.; at the same time, there are user's daily record and shopping cart information in the server. These data contains a wealth of knowledge. Through the study of user behaviour and user preferences by data mining technology, system can gain valuable knowledge, and generat personalization recommendation according to the knowledge.This thesis reseaches the current mainstream E-commerce personalized recommendation technology, collaborative filter technology, analyzes with careful consideration about the problem of sparsity and the problem of scalability, because they reduce recommended quality. It utilizes clustering algorithm to improve the collaborative filter algorithm, and invents simulation to realize the recommendation strategy mentioned in this thesis. I've done simulation on the improved algorithm. My experiment shows that the improved algorithm is much better in accuracy than traditional algorithm, it shows especial recommendation excellency in the sparse user evaluation data set. Finally an E-commerce personalization system frame is proposed. The system frame provides a reference scheme for realizing actual E-commerce personalization system.