The Study Of Link Prediction Based On Complex Networks

In recent years, the network structure and evolution mechanism has become a majorproblem in the study of complex networks. Reserchers involved in the study of complexnetworks to more and more subjects,such as physics, biology, economics, informationscience and so on. How to discover the unobserved connections or the future connections bymeas of link information, and this issue has become increasingly important research point.Itmeans given the prediction to an unknown links or links of the future.The mainstreamresearch point of link prediction is based on the similar approach, and this point will be moresystematic.According to the existing network structure, and getting information moreaccessibly and more reliably.More importantly, using this method not only more universal,butalso can avoid some complexed process by the machine parameters. For example, in scientistsco-authored network, there are two sientists have never completed a dissertation cooperation,however, they may be collaborative research in the coming years if they have a job in an sameagency. It it difficult to predict.But,we can determine this possibility of cooperation is largerthat according to the network topology which two men belong to a small circle.More scholars have proposed a variety of similarity algorithms in different angles aboutlink prediction of complex networks, and applied to real data.This paper describes the variousclassical similarity algorithms,and based on the ideology of common neighboralgorithm,taking into account the topological properties of the node itself.In the network, thenodes degrees are different, and the small degree node are more important than the greaterdegree of common neighbore node to one node, so considering the nodes similarity scores andwe can introducing one number.Therefore,this paper proposes the new algorithm whichnamed the algorithm based the degree contribution of common neighbors.In order to verifying new algorithm`s effectiveness, this paper uses four real networkdata,then obtain the topology valus and AUC calculation accuracy through Gephi softwareand Mtlab software.Comparing the accuracy of CNBD with the other similar algorithms, weconclude that CNBD is welled than other algorithms overall,and can used it in the followingcorrelated studies.