Models for Train Passenger Forecasting of Java and Sumatra

3rd International Conference on Mathematics, Science and Education 2016

数学;自然科学;教育

Sartono^1

Finance Education and Training Agency, Ministry of Finance the Republic of Indonesia, Indonesia^1

关键词: ARIMA modeling;    Box-Jenkins;    Exponential smoothing method;    Holt-Winters;    Jakarta;    Public transportation;    Railway passengers;    Regression method;   

Others  :  https://iopscience.iop.org/article/10.1088/1742-6596/824/1/012032/pdf
DOI  :  10.1088/1742-6596/824/1/012032

来源: IOP

【 摘 要 】

People tend to take public transportation to avoid high traffic, especially in Java. In Jakarta, the number of railway passengers is over than the capacity of the train at peak time. This is an opportunity as well as a challenge. If it is managed well then the company can get high profit. Otherwise, it may lead to disaster. This article discusses models for the train passengers, hence, finding the reasonable models to make a prediction overtimes. The Box-Jenkins method is occupied to develop a basic model. Then, this model is compared to models obtained using exponential smoothing method and regression method. The result shows that Holt-Winters model is better to predict for one-month, three-month, and six-month ahead for the passenger in Java. In addition, SARIMA(1,1,0)(2,0,0) is more accurate for nine-month and twelve-month oversee. On the other hand, for Sumatra passenger forecasting, SARIMA(1,1,1)(0,0,2) gives a better approximation for one-month ahead, and ARIMA model is best for three-month ahead prediction. The rest, Trend Seasonal and Liner Model has the least of RMSE to forecast for six-month, nine-month, and twelve-month ahead.