OlapGN: A multi-layered graph convolution network-based model for locating influential nodes in graph networks

Abstract

Complex networks necessitate the identification of key nodes owing to their ubiquity across the network. Traditional methodologies, such as machine learning-based and centrality-based techniques, evaluate node relevance only on network topologies or node properties. Nevertheless, both network topologies and node attributes must be considered at the time of evaluating the relevance of nodes. As a solution to this problem, this study presents OlapGN, a deep learning model that uses Graph Convolutional Networks to identify the most significant nodes in a complicated network. By integrating the two modules (deep learning and probabilistic nature), the proposed technique identifies overlapping groups and the most significant nodes within a complex network. The suggested approach determines the most significant individuals of the overlapping community after identifying their overlap. Several experiments have been conducted on actual social networks, such as VAST, Facebook, Medicine, Computer Science, and DBLP to evaluate the efficacy of the proposed model. In locating the overlapping communities and most significant nodes in heterogeneous complex networks, the proposed method has produced far better results than all other prevailing methods used for the purpose.