Evaluation of Travel Time Estimation Models with Different Inputs

摘要

Although plenty of models have been proposed in previous researches on bus travel time estimation/prediction, little attention has been paid to the effective usage of the historical travel time. In this research, comparisons on performance of historical travel time estimation model were made among different statistical values, namely the mean, the median and the maximum probability values. The research results show that, to the wholly speaking, when buses are far away from the target stop, the model with input of the mean value performs the best on the estimation accuracy and precision, while the model with input of the median value shows the best performance when buses are near the target stop. During peak periods, the model with input of the maximum probability value is superior to that with input of other estimated values. Finally, a multi-stage usage of statistical values in travel time estimation is suggested on basis of the estimation results.