Research On Community And Neighborhood Adaptive Enhanced Link Prediction Algorithm

With the development of social economy,the scale of the Internet has shown explosive growth,and various networks including social,ecological,and paper collaboration have become increasingly complex,resulting in research on complex networks.As an important subject in the research of complex networks,link prediction is committed to discovering the existing but unknown or possible future connections in the network.It is widely used in social network analysis,biological evolution,network reconstruction,personalized recommendation and other issues.The global structure of the network and the neighborhood structure of nodes are the most direct manifestations of network information.They show network characteristics from both macro and micro perspectives.As the network scale increases,a complete global structure is not easy to obtain;in addition,the neighborhood structure formed by the direct neighbors of a node cannot reflect the difference between neighbor nodes.In view of this,this article has carried out the research of community enhancement and neighborhood enhancement link prediction methods,the main contents are as follows:(1)Community Adaptive Enhanced Link Prediction method(CALP)is proposed,which uses the community structure to enhance the global information of the network.CALP introduces the community structure of the network,so that the representation learning process contains more global information,adjusts the direction of walk through the betweenness centrality of nodes,realizes adaptive walk learning,and obtains the representation vector and uses it for link prediction.(2)Neighborhood Adaptive Enhanced Link Prediction method(NALP)is proposed,which realizes the enhancement of node neighborhood information through nearest neighbors.NALP combines the natural nearest neighbor and neighborhood similarity index to define the similar distance between nodes and the similar nearest neighbor of the node.Under the adjustment of the clustering coefficient that measures the local clustering of the node,the neighborhood structure is enhanced by the similar nearest neighbor,then walk learning to obtain the representation vector and use it for link prediction.(3)Community and Neighborhood Adaptive Fusion Enhanced Link Prediction method(CNALP)is proposed,which realizes the organic combination of network global and node neighborhood information.CNALP divides community nodes,nearest neighbor nodes,and direct neighbor nodes reasonably,and expands the walk sequence according to the characteristics of the node itself.While enhancing the performance of the neighborhood and expanding the global information,it only performs one walk learning to obtain the representation vector and use it for link prediction.In this thesis,through community enhancement and neighborhood enhancement,the node representation process is improved from the expansion of the global and the neighborhood reduction,and its effectiveness is verified through experiments,and meaningful research is carried out in link prediction.