A systemic analysis of link prediction in social network

• Published in: The Artificial intelligence review Vol. 52; no. 3; pp. 1961 - 1995
• Main Authors: Haghani, Sogol, Keyvanpour, Mohammad Reza
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.10.2019; Springer; Springer Nature B.V
• Subjects: Analysis; Artificial Intelligence; Computer Science; Data mining; Machine learning; Methods; Network analysis; Social networks; Surveys; Systematic review; Link prediction; Social network; Approaches; Benefits; Challenges
• ISSN: 0269-2821; 1573-7462
• DOI: 10.1007/s10462-017-9590-2

Link prediction is an important task in data mining, which has widespread applications in social network research. Given a social network, the objective of this task is to predict future links which have not yet observed in the current state of the network. Owing to its importance, the link prediction task has received substantial attention from researchers in diverse disciplines; thus, a large number of methodologies for solving this problem have been proposed in recent decades. However, existing literatures lack a current and comprehensive analysis of existing link prediction methodologies. Couple of survey articles on link prediction are available, but they are out-dated as numerous link prediction methods have been proposed after these articles have been published. In this paper, we provide a systematic analysis of existing link prediction methodologies. Our analysis is comprehensive, it covers the earliest scoring-based methodologies and extends up to the most recent methodologies which are based on deep learning methods. We also categorize the link prediction methods based on their technical approach, and discuss the strength and weakness of various methods.