Forecasting method of long-distance rail passenger demand fluctuations by day/time to support extra train transport service planning

Abstract

A train schedule in Japan includes regular trains which is fixed on annual basis and pre-planned extra trains whose operation dates are not predetermined. Accordingly, in order to provide efficient transport services, a plan for the daily operation of extra trains must be established based on accurate passenger demand fluctuations forecasts by day/time. Therefore, we developed a method to forecast for the demand fluctuation by day/time on a certain day in the future. The method was developed by combining several fundamental waves, which are extracted by applying independent component analysis to actual ridership records with calendar information, and information on events which was held in target areas. We confirmed that the method has high accuracy by verifying its reproducibility and forecasting accuracy. In addition, the extra trains' operation planning system, which implements the forecasting method, can estimate the load factor of all trains between stations on the planning schedule based on the forecasted demand fluctuations. Then, the system can suggest an optimal extra trains' operation plan based on these values. We estimated the load factor based on forecasted demand fluctuations with the system, and we verified its accuracy. As a result, we confirmed that we can estimate the load factor with high accuracy.