Research On Collaborative Filtering Technology Based On Multi - Attribute Dynamic Weight Adjustment

With the emergence of the Internet popularization, the network of Information speed, when in the face of the vast and mad network Information, users can’t efficient to get the real useful Information to the users themselves, this means that we are in the era of a era of "Information Overload"(Information Overload). According to information overload problem, recommender system can filter out from the explosion of incremental information in accordance with user requirements information, therefore considered one of the main measure to solve the problem. In recommender systems, the personalized recommendation can not only quickly digging into the user resources required, improve the utilization of information resources, therefore, is generally considered of essential services e-commerce technology. Now is worthy more successful personalized recommendation technology of Collaborative Filtering(Collaborative Filtering recommendation technology, but in a Collaborative Filtering recommendation algorithm is widely used at the same time, there have been some defects, such as data sparse, cold start and these problems have seriously restricted the personalized recommendation results. Not only that, but now the personalized recommendation technology mostly focus on the study of user preferences, but only one-sided focus on the change of user preferences or user preferences of a particular aspect, make the personalized recommendation accuracy is low.Project has multiple perspectives, and to distinguish between different types of projects, so that the user of the project evaluation from many Angle side also reflect the characteristics of the user’s preferences. Due to the collaborative filtering algorithm based on the project is based on the similarity score to recommend project, few from the multiple perspectives analysis project user preference to its, therefore recommend accuracy is low. Is proposed in this paper, based on the weights of multi-attribute dynamic adjustment of collaborative filtering recommendation algorithm, from the Angle of multiple projects, and the variation of the user ratings of user preference, and through the right to change the value to measure the user’s preferences in the project, in order to improve project based collaborative filtering recommendation not to two or morethings user preferences leading to recommend effect is poorer, improve user satisfaction and recommended by the accuracy of the system.In this paper, the research significance is as follows:(1) Theoretical significance: through the way of the project properties for user preferences for collaborative filtering recommendation system both user preferences, and realize the personalized recommendation provides a key basis; Introducing project main attribute weights as theoretical basis, for the study of collaborative filtering recommendation based on project evaluation provides a new train of thought and direction; Set of fixed weight to dynamically adjust scores can be more accurately capture the user’s preferences and actual change, in order to achieve a more accurate recommendations.(2) the actual significance: traditional recommendation methods cannot satisfy the user’s individualized demand, the algorithm of this paper mainly concerns the influence of user preference to provide more accurate personalized recommendation, help e-commerce platform to provide more quality services; To a certain extent, improve the traditional collaborative filtering recommendation system, can adapt itself to the requirements of e-commerce applications, to improve the network shopping environment has a certain practical significance.