Document Type : Original Article
Department of GIS, Faculty of Geodesy and Geomatics Engineering, K. N. Toosi University of Technology
University of Tehran, Faculty of Geography
Department of Biomedical Engineering, and Geomatics Engineering, University of Calgary
One of the main challenges for the transportation decision-makers of megacities is to understand, model, and predict the spatiotemporal variations in a traffic flow that has changed during the COVID-19 outbreak. These fluctuations in the transportation field are related to many factors, with the most important ones attributed to the variations in using public transportation facilities. This paper evaluates the spatiotemporal trend of public transportation facilities and traffic flow before and during the COVID-19 outbreak. The main contributions of our research are: to accurately predict urban traffic congestion based on the historical traffic data using our proposed non-linear auto-regressive with external input (NARX) artificial neural networks (ANNs) model and to identify its main spatial governing factors. The proposed model is validated based on time series data of traffic flow in Tehran, the capital of Iran in October, November, and December of 2020. According to the R and RMSE values, it has been detected that there are no significant relations between the residential land use, main streets, and distance to taxi stations for the change in traffic flow before and during COVID-19. Results demonstrated that the designed time series ANN model could accurately predict spatiotemporal traffic levels. The minimum value obtained in the results for recall, precision, and F-score is less than 0.72. The maximum quantity of RMSE is 0.235, which is the correct value for the defined process.