Predicting the evolution of complex networks via local information

TL;DR

This paper introduces a new method for predicting future links in evolving networks by modeling their dynamics through local information and iterative node similarity updates, improving prediction accuracy.

Contribution

It proposes a structured-dependent index and a dynamic system model for better future link prediction in evolving networks, addressing gaps in existing methods.

Findings

The proposed index outperforms baseline methods in experiments.

The spatial-temporal position drift model effectively captures network evolution.

The approach enhances understanding of network development mechanisms.

Abstract

Almost all real-world networks are subject to constant evolution, and plenty of evolving networks have been investigated to uncover the underlying mechanisms for a deeper understanding of the organization and development of them. Compared with the rapid expansion of the empirical studies about evolution mechanisms exploration, the future links prediction methods corresponding to the evolution mechanisms are deficient. Real-world information always contain hints of what would happen next, which is also the case in the observed evolving networks. In this paper, we firstly propose a structured-dependent index to strengthen the robustness of link prediction methods. Then we treat the observed links and their timestamps in evolving networks as known information. We envision evolving networks as dynamic systems and model the evolutionary dynamics of nodes similarity. Based on the iterative updating of nodes' network position, the potential trend of evolving networks is uncovered, which improves the accuracy of future links prediction. Experiments on various real-world networks show that the proposed index performs better than baseline methods and the spatial-temporal position drift model performs well in real-world evolving networks.