Multi-attribute Matching Method Based On Tensor Decomposition

With the rapid development of information technology,the deepening of network application and the digital change of society,there gradually formed a system of the integration of people,machine and information,namely "Cyber-Physical-Social Systems",which aims to realize personalized services for users' needs of diverse services.Now,researching on the CPSS field has been highly concerned about.There chooses to provide the service with a matching manner as a research direction in CPSS space.However,the existing matching scheme is too dependent on the probability statistics model and a specific attribute of the object,ignoring the data characteristics of the object and the characteristics of multiple attributes,it is difficult to meet the needs of CPSS multi-attribute matching.Through the research we can see that the tensor representation can effectively describe the multiple attributes of the data and the intrinsic relation between these attributes and the HOSVD of the tensor realizes the extraction of high quality data.Therefore,this paper proposes a matching model based on tensor uniform representation of data and HOSVD,so as to realize multi-attribute matching based on tensor decomposition.Firstly,two kinds of data fusion techniques in the same tensor space and different tensor spaces are proposed,which realize the unified representation of data multi-attribute and the requirement of CPSS cross-space matching.Secondly,three feature matching algorithms of single attribute,general multi-attribute and multi-attribute merging are proposed for the feature matching model,and four object matching algorithms of single attribute coefficient reduction dimension,single attribute High-order BiCG,multi-attribute merging and multi-attribute High-order BiCG are proposed for object matching model.Finally,the correctness and validity of the matching model are verified by experiments,and this paper analyzes how to use the matching model to provide users with forward-looking,personalized service to meet the different needs of users.