The Research On Link Prediction Method In Social Network

In the real world, a large number of complex systems can be represented and analyzed byabstract social networks. Link prediction in social network as a key content in the study ofsocial network is widely used in various fields and has the vital significance. Besides helpingin analyzing networks with missing data, the link prediction algorithms can be used to predictthe links that may appear in the future of evolving networks. Link prediction can make a greatcontribute to analyze the evolution mechanism of complex network and understand thepotential functions and characters of network. In recent years, link prediction is becominghotspots of this area. Meanwhile, link prediction has been widely applied in the socialnetworks, information networks, technology networks, biological networks and many otherfields.Firstly, this paper studies the relevant background and theory of link prediction in socialnetwork, analyses and summarizes the research state of link prediction algorithms. On thisbasis, the article researches the link prediction algorithms based on similarity, especiallyanalyses the bottleneck factors that affect the prediction and time complexity in this kind ofalgorithm. In order to improve the overall performance of accuracy and time complexity, anew similarity measure is introduced to determine a way of expressing the proximity, and thenmultipath walk link prediction algorithm on the basis of the similarity is proposed, called MWfor short. Taking into account the topological structure of the network and using pathparameters, MW increase the related performance of link prediction in the data set of actualnetwork. Finally, taking the Coauthor ships in network science, Hi5web site and US Air asdata set, the paper attains the best experimental parameters by testing attenuation factor andthe length of paths, achieves the Common Neighbors algorithm, Adamic–Adar Indexalgorithm, Katz Index algorithm and MW algorithm respectively, and compares theirperformance.Experimental results show that the MW algorithm has higher forecast accuracycompared with the Common Neighbors algorithm and Adamic–Adar Index algorithm,meanwhile it has lower time complexity compared with the Katz Index algorithm. The MWalgorithm makes a good balance between the prediction accuracy and time complexity.