|Publication type:||Conference paper|
|Type of review:||Peer review (publication)|
|Title:||Instance based learning for estimating and predicting traffic state variables using spatio-temporal traffic patterns|
|Proceedings:||TRB 91rd Annual Meeting Compendium of Papers|
|Conference details:||91st Annual Meeting of the Transportation Research Board, Washington, USA, 22-26 January 2012|
|Publisher / Ed. Institution:||Transportation Research Board|
|Publisher / Ed. Institution:||Washington|
|Subjects:||Instance-based learning; TRB; Traffic; State estimation|
|Subject (DDC):||003: Systems|
|Abstract:||This paper describes an instance based learning method for the estimation and short term prediction of traffic variables such as travel times, based on spatio-temporal traffic patterns. Instance based learning is a data mining method to solve regression problems, where the output variable is estimated based on the most relevant observations. The proposed method can generally be used if there are historical observations of the respective variable of interest (dependent variable) and if there are relevant features available that describe the traffic situation (independent variables). This makes the method particularly useful for the estimation and prediction of occasionally observed variables (e.g. travel times from probe vehicles) based on continuously observed parameters (e.g. data from local detector stations). The method has been applied to different training data sets, including travel times derived from probe vehicle data and from a vehicle re-identification system. The main findings are: (a) the proposed method performs equally good or better than several benchmark methods that have been applied; (b) the quality of the results varies as expected with the training data - best results could be gathered for the prediction of local occupancy, travel time prediction worked better if vehicle re-identification was used instead of probe vehicles; (c) the method could be used as a backend for a traveller information systems; and (d) potentially very useful extensions are the incorporation of the states of traffic control systems and the derivation of statistical information.|
|Fulltext version:||Published version|
|License (according to publishing contract):||Licence according to publishing contract|
|Departement:||School of Engineering|
|Organisational Unit:||Institute of Data Analysis and Process Design (IDP)|
|Appears in collections:||Publikationen School of Engineering|
Files in This Item:
There are no files associated with this item.
Show full item record
Leonhardt, A., & Steiner, A. (2012). Instance based learning for estimating and predicting traffic state variables using spatio-temporal traffic patterns. TRB 91rd Annual Meeting Compendium of Papers.
Leonhardt, A. and Steiner, A. (2012) ‘Instance based learning for estimating and predicting traffic state variables using spatio-temporal traffic patterns’, in TRB 91rd Annual Meeting Compendium of Papers. Washington: Transportation Research Board.
A. Leonhardt and A. Steiner, “Instance based learning for estimating and predicting traffic state variables using spatio-temporal traffic patterns,” in TRB 91rd Annual Meeting Compendium of Papers, 2012.
Leonhardt, Axel, and Albert Steiner. “Instance Based Learning for Estimating and Predicting Traffic State Variables Using Spatio-Temporal Traffic Patterns.” TRB 91rd Annual Meeting Compendium of Papers, Transportation Research Board, 2012.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.