A Study On Dynamic Context-oritented Recommendation Methods And System

"Information Overload" is an inherent problem of internet. When user accesses wireless Internet through the mobile terminal, Information Overload becomes more apparent. At present, with the extensive application of 3G technology, mobile commerce develop more rapidly. Many enterprises have realized that the effective solution to Information Overload, and higher efficiency, higher quality information services for mobile users mean more opportunities and greater profit margins. Therefore strengthen the research of Information Overload has become an inevitable trend and profound requirement for the development of Mobile Commerce.Information Recommendation,as a important solution to Information Overload, has attracted a very high attention from Electronic Commerce, Network Economics.Sociology and other research fields, becoming an independent discipline. The application of Information Recommendation in mobile commerce to settle Information Overload is a research hotspot, but because mobile user is in a complex and unstable situation, it is more difficult to identify the instant needs of users, making the negative influence on recommendation efficiency and quality.Nowadays,the research of Information Recommendation in complex and unstable situation has been receiving more and more attentions by domestic and foreign scholars. A number of theoretical models and supporting techniques are put forward, among which there are still some spaces for improvement in theoretical methods, implementation techniques and application systems. In tracking the development trend of related subjects home and abroad, integrating the latest research achievement in multi-disciplinary fields, and taking qualitative and quantitative analysis methods as a combination, this dissertation studies the aspects of intrinsic mechanisms, essential features, decision processes and support techniques of Information Recommendation in complex and unstable situation as follows.Firstly, the concept of dynamic context is proposed, three main context integration activities are realized in dynamic context; and the expression of recommendation problem based on dynamic context is formalized, then three problem solving modes are proposed.Secondly, a dynamic context object recognition method based on context preference is proposed, the study include the modeling of dynamic context preference for user, the user segmentation based on hypergraph partitioning and the recognition of dynamic context object based on the result of user segmentation. Then a context-based collaborative algorithm is proposed based on user preference for context, research including user similarity measure, prediction of resource evaluation and so on.Thirdly, a hypergraph model of contextual items is proposed, a method to measure preference drift is constructed.Fourthly, the Information Recommendation process in the dynamic context is analysed and then a evolutionary framework of the process is constructed. On the basis of the framework, a recommendation system analysis unit is built.Fifthly, with the perspective of system analysis unit and the application of multi-agent, a cooperation mode of system is constructed, the research includes agent structure, workflow and collaborative communication style. finally the system structure and function is designedFinally, the contents and conclusions of this dissertation is summarized, and the perspective for future work is presented.