Research On Link Prediction Algorithm Based On Natural Weight And Potential Theory

In recent years,due to the continuous development of big data technology,link prediction of complex networks has received a lot of attention,and improving the accuracy of link prediction has become a key point.Link prediction refers to the mining of the existing but unknown link prediction by summarizing the known network structure information,and also includes the prediction of the evolutionary reasoning that may produce link in the future.Aiming at the current weight information of the right of link prediction algorithm considering the problem of a single,and ignored in the process of resource transfer between two nodes by the potential resources of local transmission path,and the losses of resources such as realistic problem,put forward including power prediction algorithm based on natural weight and the allocation of resources prediction algorithm based on potential theory,the main research contents are as follows.Firstly,a weighted prediction algorithm based on natural weight(WNA)is proposed to solve the problems of the single weighted method in the existing weighted link prediction algorithm,which leads to the lack of network information and inaccurate prediction accuracy.The algorithm takes two aspects of information into consideration,one is the weight information of the edge of the network,the other is the topological propertyclustering information of the whole network.An adjustable parameter is introduced to measure the relative importance between the two types of information,so that more effective information can be obtained and the prediction accuracy can be effectively improved.Secondly,in view of the problem of potential resources in the process of node information transmission and loss in the process of transmission,a resource allocation prediction algorithm based on potential theory(PRA)is proposed.In potential theory,it is mentioned that the potential energy of a node decreases by one unit for each edge it passes,which is called energy loss.This algorithm takes into account that the resource transmission process depends mainly on neighbor nodes and distinguishes the resource transmission amount of different types of nodes with parameters.At the same time,in the transmission of local path,the contribution degree of different paths is adjusted by introducing penalty factor,so that the resource information can be transmitted in an effective path,and they can be fully utilized.In the process of information transmission,the nodes will generate edge connections,thus improving the prediction accuracy.Finally,the weighted prediction algorithm based on natural weight and the resource allocation prediction algorithm based on potential theory were experimented on different types of real data sets for many times,and compared with the classical index algorithm,to verify the accuracy and effectiveness of WNA algorithm and PRA algorithm.