Paper Title
Determination The Delay Amount For Nonlinear Autoregressive Neural Network Model: Short-Term Traffic Flow Forecasting

Abstract
Today, management systems are being developed to solve the increasing traffic problem. For this purpose, Advanced Traffic Management System (ATMS) has been developed. However, these systems need models that can make predictions about future traffic conditions.With this motivation, nonlinear auto regressive neural network model (NARNET), which makes short-term traffic flow estimation, has been developed and the appropriate delay has been determined by the analysis. Traffic measurements made in Turkey D-200 highway and the counts were arranged in 15 minutes. Mean Absolute Error (MAE), MSE, RMSE and MAPE were used to determine appropriate delay values.As a result, it is understood that the increase in the delay value up to a certain point increases the model performance. However, it is determined that test errors increase rapidly if delay value increase continues. It has been found that accurate determination of the delay value in the estimation of traffic currents significantly improves the model performance. Keywords - Short-term, Traffic Flow, Forecasting, NARNET, Neural Networks, Turkey, Delay.