Scalable Graph AI Models for Dynamic Network Traffic Analysis and Optimization

Abstract

With the increasing complexity of modern network infrastructures, ensuring real-time traffic optimization and security remains a critical challenge. Traditional network traffic analysis methods struggle with scalability and adaptability to dynamic network conditions. This paper proposes Scalable Graph AI Models (SGAI-Traffic) to enhance network traffic analysis and optimization using Graph Neural Networks (GNNs), Graph Attention Networks (GATs), and Graph Convolutional Networks (GCNs). By modeling network topology as a dynamic graph structure, our approach efficiently captures spatiotemporal dependencies, identifies congestion patterns, and optimizes routing strategies in real time. The proposed model is evaluated on largescale datasets, demonstrating a 42.5% improvement in anomaly detection accuracy, a 38.7% reduction in network latency, and a 46.2% increase in routing efficiency compared to traditional methods. The findings highlight the potential of graph-based AI models for scalable, real-time, and autonomous network management.