Estimating Incident Duration

• US 20130253808A1
• Filed: 03/23/2012
• Published: 09/26/2013
• Est. Priority Date: 03/23/2012
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Wynter, Laura, He, Qing, Jintanakul, Klayut, Kamarianakis, Ioannis
• Granted Patent: US 9,111,442 B2
• Time in Patent Office: Days
• US Class Current: 701/117
• CPC Class Codes: G06Q 10/00 Administration; Management
• CPC Class Codes: G08G 1/0104 Measuring and analyzing of parameters relative to traffic conditions
• CPC Class Codes: G08G 1/0116 from roadside infrastructure, e.g. beacons
• CPC Class Codes: G08G 1/0133 for classifying traffic situation

Abstract

A method, an apparatus and an article of manufacture for incident duration prediction. The method includes obtaining incident data for at least one traffic-related incident in a selected geographic area, obtaining traffic data for the selected geographic area, spatially and temporally associating the at least one traffic-related incident with the traffic data to generate incident duration data for the at least one traffic-related incident, and predicting incident duration of at least one additional traffic-related incident based on the incident duration data for the at least one traffic-related incident.

Claims

1. A method for incident duration prediction, wherein the method comprises:

obtaining incident data for at least one traffic-related incident in a selected geographic area, wherein said incident data comprises incident type, one or more incident characteristics, and spatio-temporal information corresponding to the at least one traffic-related incident;

obtaining traffic data for the selected geographic area;

spatially and temporally associating the at least one traffic-related incident with the traffic data to generate incident duration data for the at least one traffic-related incident; and

predicting incident duration of at least one additional traffic-related incident based on the incident duration data for the at least one traffic-related incident and incident type, one or more incident characteristics, and spatio-temporal information corresponding to the at least one additional traffic-related incident;

wherein at least one of the steps is carried out by a computer device.

2. The method of claim 1, wherein predicting incident duration of at least one additional traffic-related incident based on the incident duration data for the at least one traffic-related incident comprises implementing a hybrid predictive model that combines an algorithm for recursive partitioning based on at least one permutation test with a quantile regression model.

3. The method of claim 2, wherein implementing a hybrid predictive model that combines an algorithm for recursive partitioning based on at least one permutation test with the quantile regression model comprises employing a regression tree wherein a regression model is fitted to each final node of the tree to provide predicted values of incident duration.

5. The method of claim 1, wherein obtaining incident data further comprises classifying incident types into at least one group.

6. The method of claim 5, wherein the at least one group comprises at least one of collision, disabled vehicle and traffic hazard.

7. The method of claim 1, wherein obtaining incident data for at least one traffic-related incident in the selected geographic area comprises obtaining incident data from a source providing details on incident characteristics.

8. The method of claim 1, wherein obtaining traffic data for the selected geographic area comprises obtaining traffic data from a road surveillance system.

9. The method of claim 1, wherein spatially associating the at least one traffic-related incident with the traffic data comprises matching the at least one incident with a closest traffic data record that satisfies a location description of the at least one incident.

10. The method of claim 9, further comprising:

identifying at least one upstream traffic detector and at least one downstream traffic detector from the closest traffic detector that satisfies the location description of the at least one incident.

11. The method of claim 1, wherein temporally associating the at least one traffic-related incident with the traffic data comprises applying an incident detection algorithm to trace a difference in occupancy between adjacent traffic detectors through time and to detect an incident when the difference changes.

12. The method of claim 11, wherein the incident detection algorithm associates incident data with upstream and downstream traffic data, locates a time stamp when a shockwave hits a nearest upstream detector, and records traffic data before and after the at least one incident'"'"'s time of occurrence.

13. The method of claim 1, wherein the traffic data comprises an environmental condition.

14. The method of claim 1, wherein the traffic data comprises sequential traffic measurements for the geographic area.

15. The method of claim 1, wherein the traffic data comprises at least one temporal indicator.

16. The method of claim 15, wherein the at least one temporal indicator comprises at least one of time of day and day of week.

17. The method of claim 1, wherein the incident data comprises at least one of incident severity and incident clearance time.

18. The method of claim 1, further comprising:

incorporating traffic data from before and after the at least one incident occurrence.

4. (canceled)

19. A method for incident duration prediction, wherein the method comprises:

obtaining incident data for at least one traffic-related incident, wherein said incident data comprises incident type, one or more incident characteristics, and spatio-temporal information corresponding to the at least one traffic-related incident;

predicting incident duration of at least one additional traffic-related incident based on the incident data for the at least one traffic-related incident and incident type, one or more incident characteristics, and spatio-temporal information corresponding to the at least one additional traffic-related incident by implementing a predictive model that combines an algorithm for recursive partitioning based on at least one permutation test with a quantile regression model; and

calibrating the quantile regression model without traffic data to predict the incident duration of the at least one additional traffic-related incident without traffic data;

wherein at least one of the steps is carried out by a computer device.

20. An article of manufacture comprising a computer readable storage medium having computer readable instructions tangibly embodied thereon which, when implemented, cause a computer to carry out a plurality of method steps comprising:

obtaining incident data for at least one traffic-related incident in a selected geographic area, wherein said incident data comprises incident type, one or more incident characteristics, and spatio-temporal information corresponding to the at least one traffic-related incident;

obtaining traffic data for the selected geographic area;

spatially and temporally associating the at least one traffic-related incident with the traffic data to generate incident duration data for the at least one traffic-related incident; and

predicting incident duration of at least one additional traffic-related incident based on the incident duration data for the at least one traffic-related incident and incident type, one or more incident characteristics, and spatio-temporal information corresponding to the at least one additional traffic-related incident.

21. The article of manufacture of claim 20, wherein temporally associating the at least one traffic-related incident with the traffic data comprises temporally associating the at least one traffic-related incident with the traffic data applying an incident detection algorithm to trace a difference in occupancy between adjacent traffic detectors through time and to detect an incident when the difference changes.

22. The article of manufacture of claim 20, wherein predicting incident duration of at least one additional traffic-related incident based on the incident duration data for the at least one traffic-related incident comprises implementing a hybrid predictive model that combines an algorithm for recursive partitioning based on at least one permutation test with a quantile regression model.

23. A system for incident duration prediction, comprising:

at least one distinct software module, each distinct software module being embodied on a tangible computer-readable medium;

a memory; and

at least one processor coupled to the memory and operative for;

obtaining incident data for at least one traffic-related incident in a selected geographic area, wherein said incident data comprises incident type, one or more incident characteristics, and spatio-temporal information corresponding to the at least one traffic-related incident;

obtaining traffic data for the selected geographic area;

spatially and temporally associating the at least one traffic-related incident with the traffic data to generate incident duration data for the at least one traffic-related incident; and

predicting incident duration of at least one additional traffic-related incident based on the incident duration data for the at least one traffic-related incident and incident type, one or more incident characteristics, and spatio-temporal information corresponding to the at least one additional traffic-related incident.

24. The system of claim 23, wherein the at least one processor coupled to the memory operative for temporally associating the at least one traffic-related incident with the traffic data is further operative for applying an incident detection algorithm to trace a difference in occupancy between adjacent traffic detectors through time and to detect an incident when the difference changes.

25. The system of claim 23, wherein the at least one processor coupled to the memory operative for predicting incident duration of at least one additional traffic-related incident based on the incident duration data for the at least one traffic-related incident is further operative for implementing a hybrid predictive model that combines an algorithm for recursive partitioning based on at least one permutation test with a quantile regression model.