Highlights Outliers represent potentially extraordinary patterns in traffic flow series. Outliers can be detected using residuals and predicted conditional variances. Outliers affect significantly short term traffic flow forecasting system structure. Outliers should be assimilated into short term traffic forecasting system.
Abstract Outliers in traffic flow series represent uncommon events occurring in the roadway systems and outlier detection and investigation will help to unravel the mechanism of such events. However, studies on outlier detection and investigations are fairly limited in transportation field where a vast volume of traffic condition data has been collected from traffic monitoring devices installed in many roadway systems. Based on an online algorithm that has the ability of jointly predict the level and the conditional variance of the traffic flow series, a real time outlier detection method is proposed and implemented. Using real world data collected from four regions in both the United States and the United Kingdom, it was found that outliers can be detected using the proposed detection strategy. In addition, through a comparative experimental study, it was shown that the information contained in the outliers should be assimilated into the forecasting system to enhance its ability of adapting to the changing patterns of the traffic flow series. Moreover, the investigation into the effects of outliers on the forecasting system structure showed a significant connection between the outliers and the forecasting system parameters changes. General conclusions are provided concerning the analyses with future work recommended to investigate the underlying outlier generating mechanism and outlier treatment strategy in transportation applications.
Real time traffic flow outlier detection using short-term traffic conditional variance prediction
Tramsportation Research, Part C: Emerging Technologies ; 50 ; 160-172
2014-07-29
13 pages
Article (Journal)
Electronic Resource
English
Real time traffic flow outlier detection using short-term traffic conditional variance prediction
Online Contents | 2014
|Short‐term traffic flow prediction with linear conditional Gaussian Bayesian network
Online Contents | 2016
|Short-term traffic flow prediction using time-varying Vasicek model
Elsevier | 2016
|