Research On Group Recommendation Based On User Behavior Analysis

With the rapid development of Internet technology,the number of web services has grown with an exponential rate.Users can not make a choice quickly and accurately faced with these massive services.As an effective information filtering method,recommendation system is one of the most important ways to alleviate the issue of information overload and personalization.At present,analyzing the user behavior is the most important method to recommend system to discover the user's interest.Nevertheless,the analysis of the granularity is too large and the analysis scope is too small,therefore,it is difficult to find the real-time interest of the user accurately and recommendation effect is hardly satisfactory.Additionally,in real world recommendation scenes,more and more recommend objects are transformed from individuals into groups,and traditional individual recommendation technologies for individuals also become difficult to apply.Therefore,this thesis combines user behavior analysis and group recommendation starting a research work.Main work are as follows:Firstly,this thesis proposes a group discovery method based on convolution probability matrix decomposition.This method combines content recommendation and collaborative filtering recommendation technologies.It employs description information of the service and user rating behavior and time information to analyze the user's recent interests.These user interests are clustered to generate groups.Secondly,this thesis proposes an online group recommendation method based on hybrid recommendation fusion.This method uses online learning to update users' preferences and group associations based on the Bayesian personalized ranking model and improves the existing group recommendation fusion model.Finally,in accordance with above proposed methods,a prototype system of group recommendation based on user behavior analysis is constructed and an application demonstration of a movie group recommendation system is given.The construction of the prototype system follows the steps of requirement analysis,summary design,detailed design,and accomplishes the function modules such as user group discovery,group recommendation and online feedback and verifies the feasibility and effectiveness of the proposed methods in the real scene.