Research On Link Prediction Algorithm Based On Path And Asymmetric Clustering Coefficient

Complex network is intuitively a network with high complexity like the Internet.The complex network is used to model and analyze the internal laws of the actual system,so as to find some theoretical basis and solution for practical problems.Link prediction is an important research direction.It uses some local information of nodes,network structure features contained in the path etc.to recover missing links in the network,it help us determine the overall structure in the future network.Therefore,the direction of link prediction has received extensive attention.The existing similarity algorithm is mainly based on node local attributes and path information.The algorithm based on local attribute of nodes is simple to calculate,but the accuracy is greatly affected in the case of incomplete network structure.Although the path-based algorithm contains a lot of network structure information,the computational complexity is relatively high.In this paper,the link prediction algorithm based on network local structure is proposed,it mainly proposes three related link prediction algorithms by using effective short path information and node correlation clustering coefficients.The main work and results of this paper are as follows:1.A link prediction algorithm based on node and path strength and community information is proposed.Aim at the problem that the computational complexity is not high in the case of incomplete network structure and the computational complexity of extracting high-order path information is large,This paper combines node attributes and paths,emphasize the strong influence of short path,at the end improve accuracy through community information,the rationality of the method is proved by comparing with other indicators.2.A link prediction algorithm based on adaptive degree for asymmetric link clustering is proposed.Aim at the problem that the node local attributes such as clustering coefficient cannot distinguish the contribution of two ending nodes due to their different importance,this paper shifts the focus from the node to the link,the asymmetric edge clustering coefficient is proposed,and the common neighbor degree is adaptively punished by network average clustering coefficient in the different networks,experiments show that the method effectively improves the accuracy.3.A link prediction algorithm based on node gravity is proposed.This paper applies the law of universal gravitation to social network prediction,measure quality and distance based on node importance and path length,no matter how far apart the nodes are,there is gravitation between them to promote the link.Aim at the problem that nodes with small degrees are sensitive to low clustering coefficients in the absence of sparse networks,this paper uses the degree clustering coefficients of nodes as the quality,experiments show that the algorithm based on node gravity can be effectively used for social network prediction.