The Study Of Heterogeneous Social Network Alignment Based On Network Representation Learning

In recent years,with the popularity of mobile devices and the rapid development of social networks,and many social networking applications that have emerged have also brought many indispensable social services to people.,The interaction between multiple social networks establishes a heterogeneous social network,Aligned heterogeneous social networks can identify hidden relationships in the network or anchor links between networks,The use of social network structure information for network alignment or link prediction has become a research focus.Heterogeneous network alignment has important social-economic effects on cross-network information dissemination,social links prediction,user portraits,friend recommendation,personalized social service recommendation,and online public opinion monitoring.However,the current network alignment method still has a great shortage in the face of large networks or sparse networks.Therefore,this paper proposes a network alignment method based on network representation learning.The main research contents are divided into the following sections:(1)This paper first studies the network representation learning in social networks.Representation learning,also known as express learning,the key technique in representation learning is to better represented by contextual information,thereby getting better representation.Effective representation can simplify the difficulty of dealing with problems.This paper studies and summarizes domestic and foreign relevant research used in recent years,and compares the advantages and disadvantages of these methods.(2)This paper proposes a network representation learning method based on network coarse graining,which can maximally maintain the partial and whole characteristics of social networks at the same time.The network representation method based on the coarse-grained Method borrows the idea of graphic coarse graining,coarsely granulating the network in an iterative way until the degree of nodes in the roughed network is less than a certain threshold and terminates the iteration,which can ultimately obtain coarse grained network.Then,coarse graining method combines with existing network representation methods(such as Node2Vector,Deep Walk and LINE),representing coarse-grained network.(3)This paper establishes the IAUE network alignment model based on coarse grained network representation method,BP neural network and G-S algorithm.This algorithm first performs network representation learning on the structural information of social networks through a coarse grained network representation method.Then using BP neural network training to get the mapping function of the anchor links in heterogeneous network,which gets a candidate set of anchor links between heterogeneous networks.Finally,G-S algorithm is used to improve the correctness of anchor links matching,improving the accuracy of heterogeneous network alignment.(4)The IAUE model uses Facebook dataset,Weibo,and Douban dataset as experimental data,and compares it with six methods based on node degree,MNA,MAD,CLF,PALE,and IAUE+ respectively.Adopting evaluation indicators F1 and MAP for experimental evaluation,experimental results on multiple data sets prove that the IAUE model has high performance and good generalization ability,which can accurately identify anchor links in the network.