A Comparative Analysis of Graph Neural Networks for Social Network Data Mining
Anurag Shrivastava, RVS Praveen, RaviTeja Aida, Krishna Vemuri, Srinikhil Saisatya Vemuri, Saif O. Husain
Abstract
Modern computer models are required to sift through the ever-increasing amounts of social network data in search of useful patterns and insights. Graph Neural Networks (GNNs) are a new breed of data mining tools that use graph topologies to accurately represent the interactions between nodes in a social network. In this paper, we compare and contrast Graph Convolutional Networks (GCNs), Graph Attention Networks (GATs), and GraphSAGE, three of the most recent and advanced GNN architectures. We look at how well each one performs in social network data mining tasks like community detection, link prediction, and node classification. We evaluate each model using multiple real-world social network datasets, considering factors such as accuracy, scalability, computational efficiency, and adaptability to dynamic networks. Our results demonstrate that while GCNs exhibit strong performance in community detection, GATs outperform in link prediction due to their attention mechanisms. GraphSAGE, with its inductive learning capability, proves effective in handling large-scale dynamic networks. Our goal in doing this comparison is to shed light on the relative merits of each GNN model and how well they perform in various social network mining contexts. Researchers and practitioners may use this study as a roadmap to find the best GNN models for their social network analysis projects.