Applying portfolio theory to prediction correction of train arrival times

Abstract

The application of portfolio theory to the prediction of train arrival times is shown to improve prediction accuracy. The "portfolio" comprises two correction methods based on a Wiener process: one uses history data for the current day and the other uses data for previous days. The error between the predicted and actual time is assumed to have a normal distribution. Portfolio theory is used to determine the optimal application of the two methods to the correction process. Simulation using actual data showed that the average error in the predicted arrival time was reduced to 4 s from 12 s when the timetable was dense. This error reduction will, for example, improve the efficiency of regenerative braking systems, in which the kinetic energy of an arriving (braking) train is electrically transmitted to a departing (accelerating) train.