Open AccessA new similarity-based link prediction algorithm based on combination of network topological features
Author(s) -
Hasan Saeidinezhad,
Elham Parvinnia,
Reza Boostani
Publication year - 2022
Publication title -
international journal of power electronics and drive systems/international journal of electrical and computer engineering
Language(s) - English
Resource type - Journals
eISSN - 2722-2578
pISSN - 2722-256X
DOI - 10.11591/ijece.v12i3.pp2802-2811
Subject(s) - similarity (geometry) , computer science , link (geometry) , graph , path (computing) , social network (sociolinguistics) , data mining , social network analysis , k nearest neighbors algorithm , complex network , algorithm , topology (electrical circuits) , theoretical computer science , artificial intelligence , mathematics , social media , world wide web , computer network , combinatorics , image (mathematics)
In recent years, the study of social networks and the analysis of these networks in various fields have grown significantly. One of the most widely used fields in the study of social networks is the issue of link prediction, which has recently been very popular among researchers. A link in a social network means communication between members of the network, which can include friendships, cooperation, writing a joint article or even membership in a common place such as a company or club. The main purpose of link prediction 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 paper, three new similarities, degree neighbor similarity (DNS), path neighbor similarity (PNS) and degree path neighbor Similarity (DPNS) criteria are introduced using neighbor-based and path-based similarity criteria, both of which use graph structures. The results have been tested based on area under curve (AUC) and precision criteria on datasets and it shows well the superiority of the work over the criteria that only use the neighbor or the path.