The Main Technology's Research Of Information Collaborative Filtering Based On MAS

With the rapid development of the Internet and the worldwide popularity of the Web is growing exponentially. In the unlimited, disorderly, and the limitless space, it is already becoming a great difficulty affair if people want to require lots of information, which they are longing for quickly and precisely. How to help users to get real access to the information network of information technology from the Internet massive information has become urgent to solve the problem areas. To solve this problem, intelligent information filtering technology is becoming a very important research direction. Information filtering technology development should be constantly close to the demands of users to simulate human intelligence. Intelligent, personality developing have becoming an inevitable trend of development.This paper mostly aims at the personality of network information services, analysis of the feedback information to interested users. Which is this paper's investigation information to users which no longer be interested should be filtered timely, otherwise information that may be of interest to users should be recommended. The purpose of the study is that the client browser interface introduced interface agent, adaptive learning agent, and collaborative filtering agent's teamwork to strengthen and improve the function of Client browser, and improve the quality of information collection and information search efficiency. Thereby it can realize to retrieve information of User-oriented interest. The main job includes:①Presenting User Interest Information Collaborative Filtering Prototype System based on multi-agent system. In this system, detailed designing ICF's frame, then proceeding functional description of interface agent, learning agent and collaborative filtering agent.②Presenting user's interest feedback learning algorithm based reinforcement learning algorithm. By observing current online user behavior and feedback, learning agent uses dynamic Q learning algorithm to update user model.③Improving on Frequency Pattern Tree algorithm in mining frequency pattern. In this paper, designing a kind of Interesting Weight Frequency Pattern Tree algorithm with user weighted interest. IWFP-tree information filtering algorithm deduces new interest to user by mining user interest item in other users interesting database who has similar interest to the former.④Designing ICF based on MAS, and realizing part systemic function using JACK Agent-oriented.We use Jiangsu University Digital Library capacity to nearly 100,000 words of 500 articles as a source of data for the prototype ICF to test. Experimental results show that compare to The Collaborative Filtering Systems based Frequency Pattern-Growth, listed CFS-growth and The Traditional Collaborative Filtering System, listed TCFS, ICF's recall and accuracy rate are higher than CFS-Growth and TCFS.Our innovationg has been the subject of Jiangsu University Students Fund.