Research On Context-Aware Group Service Recommendation

With the rapid development of Internet technology,and the rise of services number on the network,recommendation system is put forward.Traditional recommendation system less considers the users' context,however the improvement of accuracy and users' satisfaction of context for recommendation system has been fully validated.In addition,traditional recommendation systems are designed only for single user in most cases,and can't handle user groups in real life,other studies have suggested that single user in group can obtain good recommendation.This thesis combine the context with group recommendation to carry out the research work.The main works are as follows:First of all,from the perspective of user group discovery,this thesis mainly aims at the existing group discovery methods which ignore the user profile has time migration and group overlapping problems,it proposes a dynamic group discovery method which is based on peak density clustering method.The method firstly obtains users' dynamic tendency by decomposition of dynamic poisson,then predict users profile under different time context through the higher order singular value decomposition.It builds high similarity user set according to users' tendency,and uses the improved clustering algorithm based on density peak to implement group discovery.The simulation results show that the proposed group discovery method has better effect in group recommendation.Secondly,from the perspective of the group recommendation,this thesis proposes a group recommendation method based on core user groups and random walk with restart model(RWR-CU).It mine core groups which meet the minimum error and maximum coverage instead of the original group,then use random walk with restart model to obtain the preference relations between core user groups and different items.It uses preference fusion strategy to get group recommendation results finally.Simulation experiment results show that the proposed method reduces the computational complexity,filtering a large number of interference from user data and improve the effect of recommendation,and is helpful to the recommendation of unpopular items at the same time.Finally,based on the above theory and method,this thesis builds a prototype system based on group discovery and group recommendation,and also an catering recommend application demonstration is given.The realization of the system follows the requirements analysis,general design,detailed design and its implementation process,and also completes the user tendency access,user group discovery and group recommendation function modules.Feasibility of the proposed algorithm is verified,showing the effect of group discovery combines time context and group recommendation with core groups in real scenarios.