Hybrid dwell time prediction method for bus rapid transit based on ARIMA-SVM model

To explore a reliable dwell time prediction technology through experiments, the physical process of bus rapid transit (BRT) when it stays at the stops is analyzed. Both the longitudinal correlation and nonlinear effects from other traffic subsystems are included in this process. Therefore, the dwell time can be divided into the linear and nonlinear parts. Accordingly, autoregressive integrated moving average(ARIMA)model and support vector machine (SVM)are adopted to predict these two parts, and the final prediction results are produced by combining the two parts. Thus, the hybrid dwell time prediction method for BRT is established. The dwell time and the relative data gained at two stops in BRT Line 2 in Changzhou are modeled. The results indicate that the hybrid prediction method is effective. Compared with the single ARIMA and SVM models, the hybrid prediction method has a sharp decline of the mean absolute error (MAPE) and the mean square error (MSE). Also, the target percent whose prediction error is less than 1 s significantly increases. Furthermore, the MAPE, MSE and the target percent can reach 0.62%, 4.05 s2 and 96.79%, respectively, when training data is enough. © 2016, Editorial Department of Journal of Southeast University. All right reserved.