The Research And Application Of Dynamic Recommendation System Based On The Time Context

Recommended system as an important means to solve the information overload has caused a wide range of research and commercial applications.Earlier studies recommended system mainly from the relationship between the user and the items starting researchers evaluated through the analysis of user data and history to predict the extent of the user's favorite items,while ignoring the impact of-time information on the recommended automatic effect.In fact,the user's preferences and the types of goods are time-change,changed time information has a very important influence on the result of the recommendation system,more and more researchers began to study the dynamic binding time information recommendation system.In this article,therefore,the existing majority recommendation system do not consider time factor,causing the system to recommend the quality is not high,the gradient descent(RSVD)of the matrix decomposition algorithm was studied and improved.Based on the experimental MovieLens data set time and contextual user bias and offset information,at the same time,presents a dynamic FeatureTRSVD algorithm based on time context combined with the characteristics of user information and feature information goods.And through the experiment,the system dynamic recommended algorithm to recommend the quality and efficiency has improved.This paper mainly has the following several aspects work:First,this article introduces the overview of traditional recommendation system,and gave a brief explanation to the definition of dynamic system.In addition this paper introduces several traditional collaborative filtering recommendation algorithm,and points out the existing problems of recommendation algorithm,traditional similarity calculation and recommendation recommendation quality assessment standards.This article is to time context is presented,and the theory foundation for the research work of this article.Second,based on the gradient descent(RSVD)of the matrix decomposition algorithm is based on the research and improvement,the experimental data set time and contextual users and items bias factor,and fusion,the feature information of the users and items,put forward an improved algorithm of FeatureTRSVD dynamic.Third,this paper contains the time information of data sets in Mahout platform related experiments,the experimental results show that the proposed an improved algorithm of FeatureTRSVD dynamic has nothing to do with the time of the traditional static collaborative recommendation algorithm is compared,it is effective to improve the quality of the recommendation of the system,but also raise the efficiency of system is recommended.Finally,this paper based on the theory of dynamic recommendation system prototype system is designed.The system can real-time response to the recent behavior of users,and according to the change of user behavior to real-time adjust the recommended as a result,so as to improve the user experience in the recommendation system.