A Survey on Origin-Destination Flow Prediction

Origin-Destination (OD) prediction is a widely studied area across various disciplines, including transportation, urban planning, and computer science. This survey provides a comprehensive review of OD prediction, covering theoretical foundations, practical challenges, and advanced computational models such as machine learning and deep learning. We examine key aspects like input features, model types, datasets, and evaluation metrics, highlighting the trade-offs between utility and performance, as well as the challenges of model interpretability and scalability across different applications. The advantages of modern machine learning and deep learning models, such as their ability to handle large-scale spatiotemporal data, are contrasted with their inherent complexity and computational cost. Finally, we identify current challenges and propose future directions, such as improving model scalability, enhancing interpretability, and integrating privacy-preserving mechanisms, aiming to bridge gaps across disciplines and promote integrated solutions for OD prediction.