Research On Graph Embedding Algorithm Of Social Network User Anchor Link Prediction

In recent years,with the increasing number of social network service platforms,users can selectively participate in multiple social networks at the same time and enjoy their services.There are many types of social network application services,such as friend recommendation.In order to better implement services,it is necessary to establish a mapping relationship between virtual users in different social networks and users in the real world,that is,the problem of user anchor link prediction.Since the extraction quality of user feature attributes will affect the effect of anchor link prediction,in order to better solve the anchor link prediction problem,this paper studies the embedding algorithm,and proposes the Deep Walk embedding algorithm for sparse graphs and the SDNE algorithm based on incremental strategy.First of all,in view of the performance bottleneck problem of the classic Deep Walk embedding algorithm in processing sparse graphs,this paper proposes a Deep Walk graph embedding algorithm for sparse graphs.The algorithm first selects the nodes participating in the random walk based on the statistical analysis of the graph,and then uses the breadth-first strategy for random walk,and finally obtains the embedding vector of the node through training,and uses the embedding vector to train the user anchor link prediction model.The experimental results show that the user characteristics obtained by the Deep Walk graph embedding algorithm for sparse graphs can effectively improve the performance of the user anchor link prediction algorithm.Secondly,since most of the embedding algorithms represented by SDNE are embedding on a snapshot of the graph at a certain moment,once the social network structure changes,this type of embedding algorithm needs to re-embed the entire graph at the current moment,resulting in larger calculations Overhead.In response to the above problems,this paper proposes an SDNE embedding algorithm based on an incremental strategy.The incremental strategy is added to the SDNE algorithm.When the graph structure changes,the embedding of the graph at the previous moment is the basis.The large original node is embedded.Experiments show that the SDNE embedding algorithm based on the incremental strategy proposed in this paper can reduce the time cost of the SDNE algorithm for embedding the incremental part while maintaining the performance of the SDNE algorithm.