New Criteria for Predicting Links based on Node Composition and Network Structure

 Currently, the study of social space and social networks and the analysis of these networks has grown significantly. In real life, people are not independent of each other. People in social groups are interdependent. One of the most widely used fields in the study of social networks is the issue of link identification, which has recently become popular among domestic and foreign researchers. Link prediction can be used not only in the field of social networking, but also in areas such as bioinformatics, to explore the interrelationships between proteins, in the field of e-commerce. The main purpose in the field of link identification is to investigate the possibility of creating or deleting links between members in the future state of the network using the analysis of its current state. In this study, using local criteria of neighbor-based similarity and general path-based similarity criteria, both of which use the graph structure, new similarity criteria have been introduced. The results of the work have been tested based on precision and AUC criteria on the data set and show the superiority of the proposed method that uses a combination of graph structure information like path, neighbor and node degrees over the criteria that use only the path or the neighbor.