Approximating Network Centrality Measures Using Node Embedding and Machine Learning

Analyzing and extracting useful information from real-world complex networks has become a key challenge due to the their large sizes such networks achieve nowadays. For instance, depending on the intended node centrality, it becomes unfeasible to compute it for such large complex networks due to the high computational cost. One way to tackle this problem is by developing fast methods capable of approximating the network centralities. In this paper, we propose an approach capable of efficiently approximating node centralities for large networks using Neural Networks and Node Embedding echniques. We thus call our approach the Network Centrality Approximation using Graph Convolutional Networks (NCA-GCN) model.In contrast to recent related work, the NCA-GCN model requires only the degree centrality of each node in order to predict any other centrality. We show that the NCA-GCN model works well in different node centralities for different network sizes.