| With the rapid development of the Internet, we have already come into a new age of information overload. How to automatically filter interesting things from massive information for users has increasingly become an urgent problem to be solved. As the same time, ubiquitous network is eperiencing an age of rapid development which brings strong appeal on the personalized information service. Personalized recommender system is the method to resolve these above problems. Nowadays, how to express, calculate and update user interest model has become research hotspot of personalized recommender system.User interest model is the base component of personalized recommender system and quality of user interest modeling directly determines the quality of the personalized recommender system. The thesis studied the basic process, principle and algorithm of personalized recommender system and studied the the four stages of user interest model which is user data collection, user model expressing, user model learning and user model updating.The purpose of the thesis is mainly to complete the design and implementation of intelligent recommender engine in the key national projects "the research of key technology of multiple terminal collaborative network control platform in ubiquitous network", as well as to complete the research of the user’s interest distribution pattern. The main work and innovation points are as follows:(1) The thesis made a deep analysis of the key national projects and complete the detailed design work which include of video and document recommendation scene and service component recommendation scene.(2) The thesis developed intelligent recommender engines on the basis of the detailed design by using Java and verified the performance of the recommender system algorithms.(3) The thesis proposed a user interest distribution pattern measurement method which quotes the Gini coefficient. Result on movilens dataset verified the effectiveness of this method. |
PDF Full Text Request