Web proceedings papers


Ivan Klandev , Marta Tolevska , Kostadin Mishev and Dimitar Trajanov


Implementation of a smart parking system providing predictions about real-time parking occupancy is considered to be crucial when managing limited parking resources. In this study, we present a methodology based on machine-learning regression models for predicting parking availability. We use traffic congestion information and garage occupancy as input to the model gathered from public services, and we predict the parking availability in the same garage sixty minutes later. When using the XGBoost regression model, we achieve MSE=0.0567 which confirms the efficiency of our methodology. Additionally, we find that the times- tamp and the current parking availability value are the most influencing factors in prediction which proves the auto-regressive nature of the observed problem.


Public parking, Parking prediction, Smart city, Smart parking, Traffic congestion, Garage availability, Regressive model, Machine learning