Deep Learning for Traffic Prediction: Methods, Analysis, and Future Directions

Abstract

In an clever transportation gadget, visitors prediction is critical. Accurate visitors forecasting can assist with direction planning, car dispatching, and visitors congestion reduction. Due to the complicated and dynamic spatial-temporal relationships among one of a kind elements in the street community, this trouble is tough to solve. Recently, a considerable quantity of studies paintings has been dedicated to this area, specifically the deep mastering technique, which has extensively stepped forward visitors prediction abilities. The intention of this have a take a observe is to offer a whole evaluation of deep mastering-primarily based totally on visitors prediction algorithms from numerous angles. In particular, we offer a taxonomy and a precis of acknowledged visitors prediction algorithms. Second, we offer a listing of contemporary-day today's methodologies for numerous visitors forecast programs. Third, we acquire and set up normally used public datasets from the literature to make it simpler for different researchers. Furthermore, we adopt thorough experiments to evaluate the overall performance of various methods on a real-global public dataset to offer an assessment and analysis. Finally, we discover the field's unsolved problems.