Graph Neural Networks (GNNs) are advanced deep learning models designed for graph-structured data, capturing relationships (edges) between entities (nodes). Unlike traditional neural networks that operate on grids or sequences, GNNs work in non-Euclidean space, making them suitable for complex domains. They use message-passing algorithms to update node representations by aggregating information from neighbors, enabling both local and global pattern learning. Applications include biomedicine (protein interactions, drug discovery), recommendation systems, traffic forecasting, fraud detection, knowledge graph completion, and social network analysis. This study considers GNN architectures like GCN, GAT, and GraphSAGE as independent variables, with dependent variables including prediction accuracy, interpretability, and scalability across domains such as healthcare, e-commerce, and cybersecurity.
