Identifying Critical Nodes Based on Feature Engineering

Abstract

To mine important nodes in complex networks is very important for analyzing and governing real complex systems. Designing a good indicator that reflects the importance of nodes is a key issue on efficiently and accurately mining critical nodes. On the bases of the neighbor information of nodes, the features that can effectively reflect the local structure of nodes are extracted through feature extraction and reconstruction. The relational model between local structure and real importance of nodes is established by utilizing regression model based on the extracted features. The experimental results on 13 real networks show that the proposed method outperforms the benchmark methods of critical nodes identification.