Dedy Dwi Prastyo, Suhartono : Univariate and Multivariate Time Series Models to Forecast Train Passengers in Indonesia

Dedy Dwi Prastyo S.Si., M.Si.
Suhartono S.Si

Year

2016

Published in

3RD INTERNATIONAL CONFERENCE ON RESEARCH, IMPLEMENTATION AND EDUCATION OF MATHEMATICS AND SCIENCE

External link

Type

Seminar Internasional

Keywords

forecasting; train passengers; ARIMA; VAR; RMSE


Abstract

Time series model is one of quantitative methods that frequently used for forecasting a number of train passengers in certain route. In general, there are two types of time series models, i.e. univariate and multivariate time series. The objective of this paper is to apply ARIMA model as a univariate method and VARIMA as a multivariate method for forecasting a number of executive train passengers in Indonesia, particularly Surabaya-Jakarta route. The number of daily train passengers in three types of executive classes that departure from Surabaya Pasar Turi station, i.e. Argo Bromo Anggrek Pagi, Argo Bromo Anggrek Malam, and Sembrani, are used as case study. The data are consisted 761 observations and recorded from January 1st, 2014 till February 27th, 2016 and divided into two parts, i.e. January 1st, 2014 to January 30th, 2016 and 1-27 February 2016 as training and testing data, respectively. Root mean of squares error (RMSE) in testing data is used as criteria to select the best forecasting model. The results show that ARIMA yields more accurate forecast at two data, i.e. number of passengers at Argo Bromo Anggrek Pagi and Sembrani, whereas VARIMA gives better forecast at Argo Bromo Anggrek Malam. Hence, this result inlines with the first conclusion of M3 competition, i.e. statistically sophisticated or complex methods do not necessarily provide more accurate forecasts than simpler ones.