Publication Type

Journal Article

Version

publishedVersion

Publication Date

7-2012

Abstract

Kamarianakis et al. presented a promising application of multivariate time series analysis to the traffic condition forecasting problem in urban transportation networks. Interest is keen and growing in providing accurate predictions of link and route travel conditions, especially in the context of the burgeoning demand for ‘real-time’ traveler information. Forecasts underlying this information must be sufficiently accurate for the traveler information provided to be of tangible value. As the authors clearly pointed out, although the near ubiquitous availability of urban traffic condition data makes area-wide forecasts theoretically possible, scalability to area-wide networks and the required speed of forecast generation represent the primary challenges. As a potential answer to this two-pronged challenge, the authors proposed an autoregressive model structure that includes inputs from correlated adjacent data stations to model the dynamic deviation from a continually updated representation of the historical conditions. The method uses time-based thresholds to define distinct time periods during which separate linear models are identified and estimated using adaptive lasso regression. This discussion paper offers comments in three sections. The first section explores the relationship of the proposed method to the univariate seasonal ARIMA model presented in [1]. The second section discusses the sources of the distinct linear regimes in the proposed method. The final section offers recommendations for follow-on research that will more fully assess the utility and general applicability of the proposed method.

Discipline

Computer Engineering | Transportation

Research Areas

Intelligent Systems and Optimization

Areas of Excellence

Digital transformation

Publication

Applied Stochastic Models in Business and Industry

Volume

28

Issue

4

First Page

322

Last Page

323

ISSN

1524-1904

Identifier

10.1002/asmb.1941

Publisher

Wiley

Additional URL

https://doi.org/10.1002/asmb.1941

Download

Share

COinS