The Research And Implementation Of Social Network Alignment Based On Representation Learning

In recent years,with the vigorous development of the Internet,people have stepped into the era of "knowing the world without going out".This phenomenon is mainly due to the birth of various social platforms,such as Twitter and Wechat.These platforms are closely related to people's lives.For example,people communicate with relatives and friends through Wechat.These social softwares not only enrich people's daily life,but also provide researchers with a large number of multi domain data.In order to make full use of social networks data and integrate the user information scattered in each network,this thesis aligns the user accounts in multiple social networks.User alignment across social networks provides a feasible solution for integrating multi-source data and making full use of multiple social network information.However,social accounts are often independent of each other and lack corresponding relationships.Therefore,how to accurately identify the same user behind multiple accounts without using the supervision information has become a research hotspot.Therefore,in view of the above problems,considering the structural information,this thesis designs an effective algorithm framework to find the public space between different social networks,that is,the unified representation model.In this thesis,the alternative optimization method is designed to solve the multi constraint matrix optimization problem based on unsupervised multi network alignment and the equivalent construction is carried out.By using the idea of eigenvalue decomposition,part of the constraints are added to the solution process.A fast convergent algorithm is used to solve the matrix sub problem(common space sub problem).Finally,this thesis chose appropriate distance measurement algorithm,such as Euclidean distance,to align users.Finally,this thesis uses the real network data set for experiment,the experiment mainly includes two parts.The first part is to find the improvement of performance and effect of the user alignment algorithm designed in this thesis compared with the baseline algorithm.The second part is to set different parameters to evaluate the impact of different parameters on the algorithm designed in this thesis.Experimental results show that the algorithm designed in this thesis is better than the baseline algorithm in terms of alignment effect.