Link prediction has attracted increasing research attention recently, which aims to predict missing links in complex networks. However, the existing link prediction methods are primarily based on network structures alone, which are incapable of capturing the dynamics defined on top of the fixed network structures. In this paper, we introduce a linear dynamical response-based similarity measure between nodes into link prediction task. To address the efficiency problem, we design a new iterative procedure to avoid the explicit computation of linear dynamical response (LDR)index. Empirically, we conduct extensive experiments on real networks from various fields. The results show that LDR index leads to promising predicting performance for link prediction.
|Journal||Physica A: Statistical Mechanics and its Applications|
|State||Published - Aug 1 2019|
Bibliographical noteFunding Information:
This work was partially supported by National Natural Science Foundation of China (Grant No. 61876138 , 61672417 , 61472299 , 61602354 and 61703363 ) and Yuncheng University (China) leading discipline project under Grant No. XK-2018031 .
- Complex networks
- Linear dynamical response
- Link prediction
ASJC Scopus subject areas
- Statistical and Nonlinear Physics
- Statistics and Probability