Computational & Technology Resources
an online resource for computational,
engineering & technology publications
International Journal of Railway Technology
IJRT, Volume 3, Issue 2, 2014
A Recursive Kalman Filter Approach to Forecasting Railway Passenger Flows
M. Milenkovic and N. Bojovic
Division for Management in Railway, Rolling Stock and Traction, The Faculty of Transport and Traffic Engineering, University of Belgrade, Serbia
M. Milenkovic, N. Bojovic, "A Recursive Kalman Filter Approach to Forecasting Railway Passenger Flows", International Journal of Railway Technology, 3(2), 39-57, 2014. doi:10.4203/ijrt.3.2.3
Keywords: forecasting, railway, passenger service, SARIMA, State Space Models, Kalman filtering.
Efficient management of a railway company requires planning. In order to be effective, it is necessary to have expectations of the future conditions under which the company will operate. Forecasting represents an indispensable activity in rail transportation planning. Forecasting future demand for rail passenger services is very difficult but is necessary if companies are to succeed. This paper presents a forecasting approach for railway passenger traffic using the popular Autoregressive Integrated Moving Average (ARIMA) models in state space form. It was found that the recurring pattern of the monthly rail passenger flows is well described by a Seasonal ARIMA, that is the SARIMA (0,1,1)(0,1,1)12 model. The identified model was incorporated in the state space framework where classical Kalman recursion is applied for the calculation of forecasting values. The Kalman procedure is presented as an elegant approach for the prediction of SARIMA processes in state space form. Prediction performance of the developed SARIMA-Kalman model is compared with observed values, as well as with the outputs of simple seasonal exponential smoothing method, and demonstrate the capability and effectiveness of the proposed model in assisting managers to better predict rail passenger demand.
purchase the full-text of this paper (price £20)