Research On Cross-platform User Identification Mechanism Of Social Network

With the development and innovation of Internet technology,social platforms with different functions have gradually entered people’s field of vision.In order to enjoy different resources and services,people usually register accounts on multiple online platforms.It’s an opportunity for cross-platform research on social networks.Registered accounts scattered on different platforms and belonging to the same user build a natural bridge for information transmission and sharing between networks.Therefore,cross-network user identification has become a prerequisite for related work such as cross-platform recommendation and crossplatform user behavior analysis.If a more accurate and mature identification system can be developed,it will help promote the development of social networking applications and crossplatform technologies.According to the actual application scenarios and the nature of different platforms,we capture the correlation between the same user account on different platforms by solving the key problems such as inconsistent cross network structure,sparse location data and limitations of single type information.In order to achieve the purpose of improving the recognition performance,a link mechanism based on user interaction and spatio-temporal location records is designed and implemented.Under the premise of only using the network structure,a recognition mechanism CSRMA based on cross-network semantic representation is proposed.we propose to travel inside and outside the network from the perspectives of depth,breadth and cross-network,capture the common characteristics of the same user between networks,and solve the problem of differences in different networks;by constructing a semantic space mapping function,the influence of factors related to anchor nodes is weakened,which effectively improves the ability to identify users across networks.In addition,we deeply study the ability of spatio-temporal location records to identify users,and proposes a cross-platform social network user identification mechanism UI-STDD that integrates multiple types of spatio-temporal features.This mechanism proposes to model the bidirectional spatio-temporal dependence of position sequences based on co-attention to capture trajectory movement patterns and latent semantics,which caused data sparsity to be solved;quantify user online time distribution from different granularities to integrate personalized pattern features,resulting in improved data quality;Extract local and global features of location points,and calculate the similarity of cross-platform spatial distribution to solve the problem of space-time mismatch,which effectively optimize the ability to identify users across networks.At the same time,the algorithm integrates the three types of characteristics of spatio-temporal dependence,time preference and cross-platform distribution,makes more full use of spatiotemporal data,enhances the representation ability of user trajectories,and further alleviates the above problems.Finally,we select a number of related works for comparison on multiple cross-network relationship datasets and check-in location record datasets,and verifies through experiments that the two recognition mechanisms proposed in this paper: CSRMA and UI-STDD,can improve the ability of user identification in different application scenarios.Furthermore,this paper expands the study on the impact of combining spatio-temporal behavior and social relationship on user recognition performance.The experimental results show that the fusion of more different categories of features has better performance than single-category features.