Improving traffic prediction using congestion propagation patterns in smart cities
Attila M. Nagy, Vilmos Simon
Abstract
Accurate traffic forecast is a key task for planning transport infrastructure and real-time optimisation of traffic in large cities. The models used in professional literature usually provide accurate forecasts, but in case of congestion, forecasts can be highly inaccurate. At the heart of these situations are complex processes taking place on the road network of the city, which the prediction models are rarely prepared for. The congestion phenomena propagating on the road network of large cities have a major impact on the development of traffic patterns. In this article, we present a new traffic prediction model, the Congestion-based Traffic Prediction Model (CTPM), which refines previous forecasts based on congestion propagation patterns. Our aim is to show that using congestion data can greatly improve our forecasts. The developed model can be used in conjunction with any previous model, so there is no need to replace well-functioning methods. To the best of our knowledge, no method has yet been developed that takes traffic information into account for forecasting in such a way. Our performance studies have shown that by using CTPM we were able to refine traffic forecasts by an average of 9.76%.