A novel recovery strategy based on link prediction and hyperbolic geometry of complex networks. (10th September 2021)

• Record Type: Journal Article
• Authors: Moshiri, Mahdi; Safaei, Farshad; Samei, Zeynab
• Editors: De Domenico, Manlio
• Is Part Of: Journal of complex networks. Volume 9:Number 4(2021)
• Publication Date: 2021-09-10
• Subjects: complex networks -- link prediction -- hyperbolic geometry -- redundant link recovery -- network infrastructure failures recovery -- intentional attacks; Numerical analysis -- Periodicals; Computer networks -- Periodicals; Social networks -- Periodicals; 518.05
• DOI: 10.1093/comnet/cnab007
• Languages: English
• ISSNs: 2051-1310

Abstract

Abstract: Recovery of complex networks is an important issue that has been extensively used in various fields. Much work has been done to measure and improve the stability of complex networks during attacks. Recently, many studies have focused on the network recovery strategies after attack. In many real cases, link retrieval and recovery of critical infrastructures such as transmission network and telecommunications infrastructures are of particular importance and should be prioritized. For example, when a flood disrupts optical fibre communications in transmission networks and paralyzes the network, link retrieval corresponds to the recovery of fibre communications, so that the transmission network communication capacity can be restored at the earliest possible time. So, predicting the appropriate reserved links in a way that the network can be recovered at the lowest cost and fastest time after attacks or interruptions will be critical in a disaster. In this article, different kinds of attack strategies are provided and some retrieval strategies based on link prediction methods are proposed to recover the network after failure and attack. Beside that, a new link prediction method based on the hyperbolic geometry of the complex network is proposed to discover redundant links. The numerical simulations reveal its superiority than other common and recent link prediction-based methods used for network recovery, especially in the case of attacks based on edge betweenness strategy.