Research On Link Prediction Algorithms

Today,the research of link prediction in complex networks has attracted wide attention in the fields of physics research and computer science.As an important branch of the research on complex networks,link prediction not only helps us to understand the evolution mechanism of complex networks in theory,but also helps us to reveal relation in various disciplines.it is mainly based on a series of features such as the topology structure of the network and node attributes to predict missing data of network.Specifically,it is based on closely related factors of the network to predict the relationship between two nodes that are not connected in the network,including the prediction of edges that are already there but not yet been discovered and edges that will appear in the near future.With the advent of the big data and the development of online social platforms,existing link prediction algorithms can no satisfy the accuracy requirements of actual problems,and the accuracy of the prediction of the algorithm needs to be further improved.Link prediction based on external information on the Internet can obtain good prediction results,but in many cases it is very difficult to obtain useful information.In this paper,the existing link prediction methods are deeply studied.Based on the characteristics of network structure,the SHI algorithm based on network structure and CNBase algorithm based on common neighbor are proposed.And on this basis,through the network structure characteristics,A SHI algorithm based on network structure and a CNBase algorithm based on common neighbors are proposed.The SHI algorithm takes into account network structure holes and important nodes,We think that different types of nodes in the network have different functions,and the characteristics of each node often cannot be given by a single indicator.Through the analysis of the network structure holes and important nodes,we proposed the SHI algorithm,which has a very good effect on improving the link prediction accuracy.Experiments on real data sets show that the SHI algorithm performs better than other classical methods.This paper proposes a CNBase algorithm based on combining common neighbors and making friends.Finally,the comparison experiments show that the algorithm has betterprediction accuracy.This article proposes a link prediction method based on similarity measure,this kind of method has lower time complexity and higher prediction ability.At the same time,the similarity index that based on the network structure is not only easy to obtain but also more reliable and universal.