System and method for determining a vehicle classification from GPS tracks

1. A method for classifying a vehicle, the method comprising:

retrieving a low frequency GPS track having a sampling interval of at least 20 seconds;

deriving additional data from the low frequency GPS track, the derived additional data including interval speed and instantaneous acceleration of the vehicle;

extracting a plurality of data sets from the low frequency GPS track and the additional data;

generating a plurality of features associated with the low frequency GPS track from the plurality of extracted data sets;

establishing a kernel using an area under a curve (AUC) of a Receiver Operating Characteristics Curve (ROC);

selecting an optimal feature set for classifying the vehicle to train a machine learning model for use by a classifier,

selecting the optimal feature set by:

iteratively removing a feature below a threshold level of a vehicle classification performance from the plurality of generated features associated with the low frequency GPS track and a plurality of empirical features to create a feature set,

and

iteratively decreasing the feature set by a predetermined amount at each iteration by discarding one or more features whose removal yields AUCs below an AUC threshold level associated with classifying results from the low frequency GPS track; and

providing the optimal feature set and the kernel to the classifier, using the machine learning model, to classify the vehicle into a predetermined class of vehicles based on the low frequency GPS track.

2. The method of claim 1 , wherein the additional data further include interval acceleration of the vehicle.

3. The method of claim 1 , wherein the plurality of extracted data sets include a sequence of interval speed, a sequence of instantaneous acceleration, and a sequence of interval acceleration.

4. The method of claim 1 , wherein generating the plurality of features associated with the low frequency GPS track includes aggregating the plurality of extracted data sets by statistical functions.

5. The method of claim 1 , further comprising:

providing the plurality of empirical features to the classifier prior to selecting the optimal feature set.

6. The method of claim 5 , wherein the plurality of empirical features include a total distance of the low frequency GPS track and road types of routes covered by the low frequency GPS track.

7. The method of claim 1 , wherein the classifier includes a support vector machine having the kernel,

the kernel being a radial basis function kernel.

8. The method of claim 7 , further comprising:

collecting a plurality of training samples that have low frequency GPS tracks of vehicles with a known class; and

determining operating parameters of the support vector machine with the radial basis function kernel based on the plurality of training samples.

9. A non-transitory storage medium storing instructions, the instructions comprising:

one or more instructions which, when executed by a processor of a device, cause the processor to:

retrieve a low frequency GPS track having a sampling interval of at least 20 seconds;

derive additional data from the low frequency GPS track, the derived additional data including interval speed and instantaneous acceleration of a vehicle;

extract a plurality of data sets from the low frequency GPS track and the additional data;

generate a plurality of features associated with the low frequency GPS track from the plurality of extracted data sets;

establish a kernel using an area under a curve (AUC) of a Receive Operating Characteristics Curve (ROC);

select an optimal feature set for classifying the vehicle to train a machine learning model for use by a classifier,

select the optimal feature set, by:

iteratively removing a feature below a threshold level of a vehicle classification performance from the plurality of generated features associated with low frequency GPS track and a plurality of empirical features to create a feature set,

and

iteratively decreasing the feature set by a predetermined percent at each iteration by discarding one or more features whose removal yields atoll AUCs below an AUC threshold level associated with classifying results from the low frequency GPS track; and

provide the optimal feature set and the kernel to the classifier, using the machine learning model, to classify the vehicle into a predetermined class of vehicles based on the low frequency GPS track.

10. The non-transitory storage medium of claim 9 , wherein the additional data further include interval acceleration of the vehicle.

11. The non-transitory storage medium of claim 9 , wherein the plurality of extracted data sets include a sequence of interval speed, a sequence of instantaneous acceleration, and a sequence of interval acceleration.

12. The non-transitory storage medium of claim 9 , wherein the one or more instructions, that cause the processor to generate the plurality of features associated with the low frequency GPS track, cause the processor to:

aggregate the plurality of extracted data sets by statistical functions.

13. The non-transitory storage medium of claim 9 , wherein the instructions when executed by the processor, further cause the processor to:

provide the plurality of empirical features to the classifier prior to selecting the optimal feature set.

14. The non-transitory storage medium of claim 13 , wherein the plurality of empirical features include a total distance of the low frequency GPS track and road types of routes covered by the low frequency GPS track.

15. The non-transitory storage medium of claim 9 , wherein the classifier includes a support vector machine having the kernel,

the kernel being a radial basis function kernel.

16. The non-transitory storage medium of claim 15 , wherein the instructions when executed by the processor, further cause the processor to:

collect a plurality of training samples that have low frequency GPS tracks of vehicles with a known class; and

determine operating parameters of the support vector machine with the radial basis function kernel based on the training samples.

17. The method of claim 1 , wherein the plurality of empirical features are obtained from a third party vendor based on position information of the low frequency GPS track.

18. The method of claim 1 , where generating the plurality of features comprises:

generating the plurality of features by applying one or more statistic functions to the plurality of extracted data sets.

19. The non-transitory storage medium of claim 13 , wherein the plurality of empirical features are obtained from a third party vendor based on position information of the low frequency GPS track.

20. The non-transitory storage medium of claim 13 , wherein the one or more instructions, that cause the processor to generate the plurality of features of the low frequency GPS track, cause the processor to:

generate the plurality of features of the low frequency GPS track by applying one or more statistic functions to the plurality of extracted data sets.