Application of Multiplicative Decomposition and Exponential Smoothing Techniques for Bus Arrival Time Prediction

Informing public transit users about the expected arrival time of the next bus is one of the major components of Advanced Public Transportation Systems (APTS), which is expected to attract more travelers to public transit thereby reducing congestion on the urban roads. Most of the existing studies on bus arrival time prediction use data driven techniques such as regression, time series, neural network etc. or model based techniques like Kalman filtering. The studies which are based on time series methods for bus arrival time prediction apply Box-Jenkins models such as Autoregressive (AR), Moving average (MA) or combination of AR and MA models (ARMA, ARIMA etc.). There are classical methods in time series analysis like multiplicative decomposition, exponential smoothing etc. developed long before the Box-Jenkins models and are still the most popular forecasting methods used in business and industry. However, there are not many reported studies which explored the usefulness of the decomposition and smoothing techniques for the problem of bus arrival time prediction. The present study is one of the first attempts in this direction, by applying a multiplicative decomposition and simple exponential smoothing technique for bus arrival time prediction under heterogeneous traffic conditions. The performance of both the methods were corroborated using 105 actual bus trips data spanned over 5 days in a study route representative of a typical urban environment. The results show that, using limited bus trip data as input for both the methods, the smoothing method outperform the decomposition method with an average mean absolute percentage error (MAPE) in the range of 10-12. The major advantages of the classical time series methods are that they are simple to understand and easy to implement when compared to complex ARIMA models.

Language

  • English

Media Info

  • Media Type: Digital/other
  • Features: Figures; References; Tables;
  • Pagination: 12p
  • Monograph Title: TRB 91st Annual Meeting Compendium of Papers DVD

Subject/Index Terms

Filing Info

  • Accession Number: 01372693
  • Record Type: Publication
  • Report/Paper Numbers: 12-1880
  • Files: TRIS, TRB
  • Created Date: Feb 8 2012 5:06PM