A Hybrid Model for Short-term Prediction of Times Series Data: Forecasting Monthly Throughput for Hong Kong Port

Short-term demand prediction enjoys a critical position in the planning of business operations, especially in the logistics industry that needs to prepare sufficient capacity to satisfy the immediate future demand. Conventionally, short-term prediction is done by ad hoc time series analysis, such as auto regressive moving average for stationary series, or auto regressive integrated moving average for nonstationary series. One of the underlining assumptions in using the estimated coefficients from the time-series model for forecasting is that there are no structural changes between the time for model estimation and that for prediction. If there are any new changes in the data generation process, it cannot be reflected in the prediction. In addition, most of the predictions are satisfied if the estimator is BLUE. If the underlining assumption about data generation mechanism cannot be satisfied, the prediction accuracy cannot be guaranteed. The Kalman filter is an efficient recursive filter that estimates the states of a linear dynamic system from a series of noisy observations. It has been used in a wide range of engineering and econometric applications. It takes the least square estimator of a state vector from a previous time step, uses the current observation to update the new estimator, and then uses the updated estimator in the forecasting step. With the Kalman filter, more weight is being given to the estimates with lower uncertainty. Therefore, once the initial state of the model is established, the Kalman filter will provide better forecasting results based on the updated observation. In this paper, the authors develop a hybrid method that integrates time series analysis with the Kalman filter, to increase the accuracy of the short-term prediction. The developed model is applied in the case study for monthly forecasts of container ports in Hong Kong. The result shows that the hybrid model can significantly increase the prediction accuracy.

• Record URL:
  http://www.icms.polyu.edu.hk/Proceedings/Proceedings%20of%20IFSPA%202012.pdf
• 
• Availability:
  • Find a library where document is available. Order URL: http://worldcat.org/isbn/9789623677578
• Supplemental Notes:
  • Abstract reprinted with permission of Hong Kong Polytechnic University.
• Corporate Authors:

  Hong Kong Polytechnic University

  Department of Logistics and Maritime Studies
  Tung International Centre for Maritime Studies
  Hong Kong,   China 
• Authors:
  • Luo, Meifeng
  • Ren, Shuyun
• Conference:
  • International Forum on Shipping, Ports and Airports (IFSPA) 2012: Transport Logistics for Sustainable Growth at a New Level
  • Location: Hong Kong , China
  • Date: 2012-5-27 to 2012-5-30
• Publication Date: 2012

Language

• English

Media Info

• Media Type: Digital/other
• Features: Figures; References; Tables;
• Pagination: pp 520-527
• Monograph Title: Proceedings of the International Forum on Shipping, Ports and Airports (IFSPA) 2012: Transport Logistics for Sustainable Growth at a New Level

Subject/Index Terms

• TRT Terms: Container terminals; Container traffic; Containerships; Forecasting; Kalman filtering; Mathematical prediction; Ports
• Uncontrolled Terms: Throughput
• Geographic Terms: Hong Kong (China)
• Subject Areas: Freight Transportation; Marine Transportation; Planning and Forecasting; Terminals and Facilities; I72: Traffic and Transport Planning;

Filing Info

• Accession Number: 01485372
• Record Type: Publication
• ISBN: 9789623677578
• Files: TRIS
• Created Date: Jul 1 2013 9:55AM