Prediction of Speed on Two-way Two-lane Expressways

Abstract

This study proposes a hierarchical regression model to predict spot speed on Two-way Two-lane (TWTL) expressways. The model has two explanatory variables: longitudinal gradient and the distance from the start point of TWTL. The values of explanatory variables are estimated every rank of traffic flow rate and every TWTL route. To take heterogeneity of the effect of traffic flow rate rank and route on spot speed into account, a hierarchical model is assumed. Speed profile data on 4 TWTL routes in Japan obtained from ETC2.0 probe data are used for calibration and cross-validation; data on inbound direction is for calibration of the model while that of outbound is for cross-validation. As a result of the calibration, the root mean square error (RMSE) between the actual speed and estimated speed is lower than 6 km/h under 600-800pcu/h for section of calibration. The RMSE on cross-validation is also lower than 6km/h. There are some other findings in the analyzation; for instance, the coefficient of longitudinal gradient becomes large when traffic flow rate increases.