Research On Virtual Identity Mapping Algorithm For Network Sapce Entity

With the popularity of social networks,visiting social networks have become people's daily routine.Researches related to social networks are developing as well.However,as single social network is functionally limited and hard to meet people's different kinds of needs,people tend to register accounts on different social networks to enjoy the services that social network platforms provide.But because the information between social networks are quite different and there is no effective way ot build links between virtual identities,only part of information can be obtained in a single social network and it's hard to get integrated information on different social networks,a scientific method is urgently needed,different virtual identities need to be linked together.Accurate virtual identity mapping can support many scientific research,such as depicting complete user portraits and achieving accurate friend recommendation.In summary,the research of virtual identity mapping has a wide range of research value and is a very meaningful topic.At present,the research on virtual identity mapping mainly focuses on the attributes of virtual identity,that is,the personal information that users fill in when registering accounts.However,due to the differences between social platforms,user attributes of different platforms rarely overlap,and mapping methods based on multiple user attributes are often limited to a few social platforms.On the other hand,the mapping method based on single attribute has good applicability because it only focus on the single attribute coexisting in most platforms.However,due to the limited recognition ability of single attribute feature to users,the algorithm not only improves the scalability of application,but also sacrifices the accuracy of recognition.Aiming at the problem of "accuracy" and "breadth" of virtual identity attributes in social networks,this paper classifies virtual identity attributes in social networks,calculates the similarity degree of various attributes respectively,and proposes an entity tuple similarity algorithm based on sorting to solve the similarity calculation problem of semi-structured attributes in social networks,which effectively improves the computational efficiency.In this paper,the similarity calculation results of all kinds of attributes are applied to four classification models: logistic regression,support vector machine,decision tree and random forest.Finally,the decision tree is selected as the classification model for identifying virtual identity through model comparison experiments.Finally,the effectiveness of the proposed algorithm is proved by comparing with the two mapping algorithms.