| As the the development of information technology, e-commerce sites are springing up over the recent. Today, the Internet has brought people into the information society and network economy era. Electronic Commerce has produced a huge and far-reaching impact on enterprise and daily lives of the people’s. On the one hand people can stay at home shopping, looking at the news, watching movies. On the other hand, in order to find the information which they are interested in, users need to visit a lot of irrelevant information or products in the vast ocean of information. The people was lossed continues who was submerged in the problem of information overload.In order to effectively solve these problems personalized recommendation system come into being.The recommendation build users’interest model based on the information in the user access and project model. Then system filters and selects information from the complicated information by user model, then recommends user the project which is the interesting item for people. Clearing the preferences of users (consumers) is the recommended core. Personalized recommendation system can be seen as the following three parts:the supply part, the demand part, the supply and demand matching. The supply part is the recommended objects, such as goods, information, books, and movies. The demand part is the people who needed the recommend information. Supply and demand matching is the ties between the supply part and the demand part. The existing research focuses on the supply and demand matching part. Modeling of user interest mainly depends on the recommended method. Research on the supply part is usually whether to buy, browse, and evaluation of this project. From a system point of view, in order to make the system play a greater role, the various elements must fully play the role and organically combinate together. The lack of supply part may be the main reasons of personalized recommendation system can not achieve the accuracy of real-time; and the reason of reducing user satisfaction for the personalized recommendation system.In this paper, using the commoditiesthe basic information and the sort of users for goods combined with multi-attribute decision theory to determine the user interest model. A user preference (weights) is a convex set, when commodities sort as a known. Solving the convex set is an NP-hard problem. For the NP-hard problem, this paper put forward the method of solving a number of possible solutions, which is named as the reverse weighting. This paper also presents the use preferences analysis method based on the goods sorted and commodities basic informations for product information for text or image, and verifies the reliability and validity of the method. Finally, this method is applied to the film industry to establish a movie recommendation application. Build the movie recommendation system architecture, User behavior recording module, the recommendation algorithm application modules, and recommended the resulting output module. Through the user’s access to information and user feedback, respectively, to obtain the application of the film sort presented in this paper are two ways to find the user preferences. And apply a simple weighted weight sort of alternative film on user preferences; select TOP-N items recommended to the user. The main work is as follows:1) Based on the products sort and commodity inforamtions the reverse-weighted analysis method is proposed. And the basic idea and method step is described.2) Based on user preferences of the goods sorted the basic idea and method steps of the analytical methods are proposed.3) Analysis reliability and validity of the methods in the case of interest offset and no interest offset.4) Based on the above method to establish the movie recommendation model... |
