Optimization of Cycle Length with Short Term Traffic Prediction and Delay Analysis

Abstract

In the study, the effect of parametric and nonparametric methods on short-term traffic prediction and intersection cycle length and performance were investigated. According to the data of the intersection, it is aimed to improve the delay times and reduce the intersection waiting times and improve the intersection performance. The methods were applied, for example, to the Mimar Sinan intersection of Denizli. The data obtained with the help of sensors located in the approach arms of the intersection are arranged as data sets. Short-term traffic prediction has been made with auto-regressive integrated moving average (ARIMA) and artificial neural networks (ANN) methods. Estimation results the intersection cycle length optimization was made using the Webster method. After calculating the optimum cycle length and green times, the lag values of the Webster delay method and intersection approach arms and the intersection general were compared with the estimated results obtained from both the ARIMA method and the ANN method. In the short-term traffic prediction, the results obtained using the ANN method were found to be more successful than the results obtained with the ARIMA method.