Efficient Traffic Forecasting on Large-Scale Road Network by Regularized Adaptive Graph Convolution
Researchers have developed a new method called Regularized Adaptive Graph Convolution (RAGC) to improve the efficiency of traffic forecasting on large road networks. The model utilizes an Efficient Cosine Operator (ECO) for graph convolution with linear time complexity, addressing the scalability issues of traditional methods. RAGC also incorporates a framework that combines Stochastic Shared Embedding (SSE) and adaptive graph convolution to enhance prediction accuracy while maintaining computational efficiency. Experiments on four real-world datasets demonstrated that RAGC outperforms existing state-of-the-art methods in accuracy and shows competitive computational performance. AI
IMPACT Introduces a more scalable and accurate approach for traffic prediction, potentially benefiting urban planning and navigation systems.