Research On User Preference Extraction Based On Multidimensional Context In Mobile Internet

The rapid developing mobile Internet has gradually penetrated into every corner of people's life,bringing unprecedented transformation,reconstruction or subversion to the traditional industry.With the inherent ability of context awareness,the mobile Internet gradually creates a new situation of personalized service,which makes the pervasive computing ineffective,and brings new challenges to the traditional user preference extraction technology.Traditional user preference extraction technology focuses on the relationship between users and items,while seldom considering the influence of context information on user preference.Though,there are some personalized service researches based on a few context factors,these user preference extraction algorithms have limitations-the context factors are relatively single,which makes it difficult to extend to the multidimensional context situation.Under this background,this thesis mainly focuses on the extraction technology of user preference under multidimensional context.The main contributions are described as follows:Firstly,this thesis investigates the current research situation of user preference.User preference modelling methods and expressing methods are summarized.Besides,the extraction of user preference under multidimensional context is assessed.Secondly,based on multidimensional context factors,this thesis proposes a multidimensional context-aware user preference model.This model can adapt to the change of context factors.Moreover,using this algorithm,the user's preference can be easily calculated while multidimensional contexts are given.When constructing this model,this thesis firstly disintegrate user preference into long-term preference and short-term preference according to psychological research,and then multidimensional context factors are classified as user context,item context and environment context.Based on above,this thesis also gives an analysis of the relationship between user preference and all kinds of context factors to construct CB-MF(Context Bias Matrix Factorization)model Then,the proposed model and algorithm are simulated on LDOS-CoMoDa dataset.Experimental results show that this model improves the performance compared with the traditional MF(Matrix Factorization)algorithm,and its RMSE(Root Mean Square Error)reduces by 10.38%while MAE(Mean Absolute Error)reduces by 11.51%.At the end,standing on the result of theoretical research above,this thesis designs a video player platform based on the user preference extraction technology under multidimensional context.In addition to playing video,the web platform has other functions like collecting users'context,extracting users' preference and making recommendation.Moreover,the engine of user preference extraction is equipped with CB-MF algorithm which sufficiently demonstrates the feasibility of CB-MF algorithm.This thesis also presents details about the system structure and the implement of every modules.The platform has important meaning to the ffuture research on the relationship between multidimensional context and user preference.