Development of Empirical Models to Predict Gap Acceptance Parameter Based on the Geometrical and Operational Parameters of Different Roundabouts

This paper develops a mathematical formula that accounts the influence of roundabouts design and performance parameters to predict the gap acceptance parameter "Critical gap". Thirteen roundabouts in Hungary having different geometric and operational parameters were selected. The geometrical and operational data of each roundabouts leg was collected. Raff's method was used to estimate the critical gap for each roundabout leg. Firstly, the collinearity analysis was carried out to identify independent parameters to avoid any possible negative impact on the developed predictive models. Nine out of ten parameters passed the collinearity test. These nine parameters are the main parameters used in the model development. Three models were developed. The first model (M1) is based on Multivariate Adaptive Regression Spline (MARS) algorithm. The second model (M2) is based on Pearson correlation. The last model (M3) was based on Spearman correlation. Linear regression models were constructed using the retained parameters of the M2 and M3 models. Subsequently a comparison of the three developed models is done based on R2 and RMSE values. Based on the results obtained from the comparison, the MARS model (M1) is the best predictive model of the critical gap. According to the results of the MARS model, the most important parameters for predicting critical gap value are circulating traffic flow and the distance between neighboring legs.