Research On Multi-agent Recommendation System Based On CBR And MADM

With the popularity of computer and the World Wide Web,capturing information and purchasing products through the web has become the mainstream. However, the information resources on the web grow explosively, and people have to spend a lot of time and effort to obtain relevant information and filter the products before buying a product, so"rich data, poor information"becomes a big issue in e-commerce and web service. In order to resolve this issue, researchers have proposed many strategies and solutions, and recommendation system is one of them.However, in the existing e-commerce recommendation system, there still exists low recommendation quality because of data sparseness, single recommendation algorithm (based on the collaborative filtering and content-based recommendation), and they only consider the situation that the product property values are either completely accurate or inaccurate. This thesis addresses the above problems and examing e-commerce recommendationsystem in depth, more specifically:1. Applying agent technology to the recommendation system, the system gives full play to their intelligence, reactivity, proactivity and provides powerful technical support to information and products resource discovery and recommendation. A new multi-agent recommendation system architecture—CMARS (CBR-based Multi-Attribute Multi-Agent Recommendation System) is proposed, and its function, structure and workflow are designed. CMARS consists of five kinds of agents whose functions are independent of each other, and they divide the work to complete the recommendation task, which overcome the limitation of the traditional recommendation system.2. Applying CBR (Case Based Reasoning) method to the recommendation system, this thesis improved the problem caused by data sparsity. The case representation method is designed, which can integratedly and effectively express the user's characters and their various needs of the buying information.3. About the algorithm, a hybrid data similarity measure algorithm and TOPSIS multi-attribute decision making method based on distance are designed taking into account the property classification. In this algorithm, the customers'needs information can be either accurate or inaccurate, which is of more practical significance.4. Based on above techniques and algorithm, a web teaching material recommendation system is implemented, and experiments are designed to test the above-mentioned two algorithms. The experiments show that both the hybrid data similarity measure algorithm and the TOPSIS multi-attribute decision making method are of higher recommendation precision.In short, this thesis integrates CBR method, MADM method and Multi-agent technology together and applies them to the e-commerce recommendation system. The hybrid data similarity measure algorithm and TOPSIS multi-attribute decision making method are good attempts. The study approaches of this thesis can facilitate the research and development of e-commerce recommendation system and web service.