Personalized Service Recommendation System Research

With the rapid development of Internet technology, the application of electroniccommerce unceasingly deepen. E-commerce provides more choices for the usersprofiting from the further expansion of its scale, at the same time, its structure hasbecome more complex. People often feel lost and can’t grasp the key points whenthey face massive information, which indicates the emergence of needs toe-commerce personalized recommender systems.The personalized recommendationcan not only quickly help customers find right goods in various complicatedinformation,but also help customers make decision by comparing theinformation.There are many problems in the existing recommendation systems,suchas lack of personality, insufficient association and weak real-time.This thesis studies an e-commerce personalization recommendation systembased on collaborative filtering. First, we analyze the present situation of theelectronic commerce recommendation system research. Second, the requirementanalysis of such a system is proposed. Then, we design and develop an e-commercepersonalization recommendation system focusing on the collaborative filteringalgorithm. Finally, we detailedly test our algorithm and system. The results of testshows that the response of system meets the requirement of relative standers andprovides better service quality on sparse evaluation of user data set benefiting fromthe improved algorithm.Most work of our research are:Impoving the traditional collaborative filtering algorithm focusing on the shortof sparse and expand problem which lead to the inaccurate recommending and badreal-time, we propose a collaborative filtering algorithm and realize therecommending stratigies of our system based on concept hierarchy.This algorithm implements web data mining through analyzing the Web log,user registration information,order information and Cookies which are got fromservers.Further more, the algorithm collects the user’s scoring data to establish themodle of user recessive interest, create the database of user characteristic,implement the personalized filtering. We design and realize an e-commerce personalization recommendation systembased on the improved algorithm, analyze and design the frame and workflow of thee-commerce personalization recommendation system, and test the functions andpreferences of our algorithm.The result of experiment shows, the impoved filtering algorithm has betterrecommanding accurate, association, real-time featuers,especially reveals eminentrecommanding ablilities on sparse evaluation of user data set.