Research And Application Of Link Prediction Based On Node Similarity In Signed Networks

With the rapid development of information technology represented by the Internet,social network analysis has become a hot research field.Most of the researches on social networks analysis mainly focus on the non-signed networks,and this paper focuses on the social networks with positive and negative sign attributes,namely the signed network.As an important issue in the research of signed networks,link prediction is helpful to reveal the structural evolution of signed networks,and has a wide range of application value in the fields of user decision behavior prediction,recommendation system and protein interaction analysis.This paper focuses on the research of link prediction in symbolic networks.The main work and innovations are as follows:Firstly,most prediction algorithms in signed networks can only predict the sign of existing links with missing sign type,and they can't achieve the prediction of unknown links and their signs.In order to achieve the dual goals of link prediction and sign prediction in signed social networks quickly and accurately,and reduce the computational complexity while achieving high accuracy,a signed network link prediction algorithm TLG is proposed,which integrates local and global structural features to define the similarity between nodes.Based on structural balance theory,the algorithm defines the local link tightness and global tightness respectively by using the structure information of paths with the step size of 2 and 3connecting the two nodes.The total similarity of the two nodes is gotten,where the absolute value measures the possibility of establishing a link,and its sign is the sign prediction result of the link,so as to realize the link prediction and sign prediction in the signed networks.Secondly,aiming at the unbalanced proportion of positive and negative links in signed networks,in order to reduce the computational complexity and improve the accuracy of sign prediction,this paper proposes a signed network link prediction algorithm CNCC?SI which combines the clustering coefficient of common neighbor nodes and the influence of the sign of edges.By introducing the concept of sign influence of path L based on structure balance ring,and effectively using the information of node degree,clustering coefficient,intermediate transmission node on the path,edge connecting sign and so on,the similarity of two nodes based on first-order common neighbor and second-order common neighbor is defined respectively.Finally,the total similarity score of two nodes is obtained,and its absolute value is used to measure the possibility of two nodes establishing links In order to achieve link prediction in signed networks,the sign prediction results of links are obtained through its sign.Finally,the prediction accuracy is further improved by analyzing the influence factors of adjustable step size.The correctness and effectiveness of the proposed algorithm are verified by experiments,and the recommendation results of top K positive and negative links on signed network dataset are given.