Cross-network User Identification Based On Spatiotemporal Perception

The rapid development of the Internet and the popularization of mobile smart devices make social networks become indispensable social means of information society,and the research on social networks has become a hot research topic for academic circles and Internet enterprises.In real life,it is common for the same natural person to have different social networks,and it is of great significance for the analysis and application of social network-based to effectively identify the same user account in different social networks through the cross-network user identification method.In recent years,researchers have proposed new methods for identifying users across social networks based on the similarity of user tracks,but most of them do not take into account the strong correlation characteristics of time and space,resulting in low recognition accuracy.Therefore,based on the characteristics of time and space strong correlation,this paper designs a user recognition model across social networks,and implements the relevant algorithm of user identification UIDwST(Identification User base on spatio-Templ perception).The model first pre-processes the obtained raw data,which solves the problems of excessive noise,missing values and data dispersion.Preprocessing involves cleaning,extracting,converting latitude and longitude coordinates of positional entities,and so on.Track similarity is then calculated across social network users.According to the idea of TFIDF algorithm,the calculation of assigning different weights to different sign-in records is completed,so as to enhance the identification ability of different check-in records.The user-to-similarity calculation method which takes into account the characteristics of space-time strong correlation is designed in combination with the SIGMOID function,Finally,cross-social network user recognition is completed.Based on the similarity values calculated by all user pairs calculated by the above method,the results are sorted from highest to lowest,and the results are used as input for user similarity determinations.A user-to-judge principle-to-order user pair is established to filter and generate a"candidate-to-user collection".Using the SIGMOID function,The ReLU(the activation function commonly used in artificial neural networks)and the step function to determine the"candidate user pair",the similarity above the threshold of the user pair is determined to be the same user,and finally get the cross-network user recognition results.In this paper,the design of user identification methods on three real data sets carried out experiments and results analysis,the experimental results show that the method is feasible.The results of the experiment were then compared with three existing approximation methods,and the results showed that The accuracy obtained by the method on the three real data sets reached 0.7847,0.8528 and 0.8594,respectively,higher than the comparison experiment of 0.6984,0.8351 and 0.8580,and increased the recognition accuracy by 8.63%and 1.77%respectively,and 0.14%.