Systems and methods for driver scoring with machine learning
View Patent ↗Systems and methods for using machine learning classifiers to identify anomalous driving behavior in vehicle driver data obtained from vehicle telematics devices are provided. In one example, a vehicle telematics device receives vehicle driver data from sensors, identifies anomalies in the vehicle driver data by using an unsupervised machine learning process, calculates a driver risk score by using the anomalies identified in the vehicle driver data, and transmits the risk score to a remote server system. In another example, a server system receives vehicle driver data from a plurality of vehicle telematics devices, identifies anomalies in the vehicle driver data by using an unsupervised machine learning process, and calculates a driver risk score by using the anomalies identified in the vehicle driver data.
1. A vehicle telematics device, comprising:
a processor;
a communications device coupled to the processor;
one or more sensor devices coupled to the processor; and
a memory coupled to the processor;
wherein the vehicle telematics device:
receives a set of unstructured vehicle driver data from the one or more sensor devices;
identifies anomalies in the set of unstructured vehicle driver data by using an unsupervised machine learning process that identifies relationships in the unstructured vehicle driver data;
calculates a driver risk score by using the anomalies identified in the set of unstructured vehicle driver data and the identified relationships in the uncategorized vehicle driver data; and
transmits the driver risk score to a remote server system by using the communications device;
wherein using the unsupervised machine learning process further comprises generating a plurality of isolation forests that distinguish clusters of the set of uncategorized vehicle driver data from anomalies in the set of uncategorized vehicle driver data.
2. The vehicle telematics device of claim 1 , wherein the set of unstructured vehicle driver data is selected from a group consisting of vehicle speed, vehicle acceleration, vehicle deceleration, and vehicle swerving.
3. The vehicle telematics device of claim 1 , wherein each isolation forest in the plurality of isolation forests comprises a plurality of isolation trees.
4. The vehicle telematics device of claim 1 , wherein an isolation forest in the plurality of isolation forests without identified anomalies is labeled SAFE to indicate safe driving and wherein an isolation forest with identified anomalies is labeled UNSAFE to indicate unsafe driving.
5. The vehicle telematics device of claim 1 , wherein the vehicle telematics device calculates the driver risk score by using an ensemble scoring process.
6. The vehicle telematics device of claim 5 , wherein calculating the driver risk score by using an ensemble scoring process is evaluated by the vehicle telematics device by using the following expression:
Risk
Score
(
v
i
)
=
100
×
Num_UNUSAFE
n
where v i is a vehicle with a specific set of driver data, n is a total number of isolation forests in the plurality of isolation forests, and Num_UNSAFE is the number of isolation forests in the plurality of isolation forests labeled UNSAFE.
7. The vehicle telematics device of claim 6 , wherein the driver risk score above a predetermined value indicates an unsafe driver.
8. The vehicle telematics device of claim 1 , wherein the vehicle telematics device further calculates the driver risk score for a specific date range.
9. The vehicle telematics device of claim 1 , wherein the communications device is a wireless device.
10. A method for driver risk scoring, the method comprising:
receiving a set of unstructured vehicle driver data from one or more sensor devices by using a vehicle telematics device, wherein the vehicle telematics device comprises a processor, a memory coupled to the processor, a communications device coupled to the processor, and the one or more sensor devices coupled to the processor;
identifying, using the vehicle telematics device, anomalies in the set of unstructured vehicle driver data by using an unsupervised machine learning process that identifies relationships in the unstructured vehicle driver data;
calculating, using the vehicle telematics device, a driver risk score by using the anomalies identified in the set of unstructured vehicle driver data and the identified relationships in the unstructured vehicle driver data; and
transmitting the driver risk score to a remote server system by using the communications device;
wherein using the unsupervised machine learning process further comprises generating a plurality of isolation forests that distinguish clusters of the set of uncategorized vehicle driver data from anomalies in the set of uncategorized vehicle driver data by using the vehicle telematics device.
11. The method of claim 10 , wherein the set of vehicle driver data is selected from the group consisting of vehicle speed, vehicle acceleration, vehicle deceleration, and vehicle swerving.
12. The method of claim 10 , wherein each isolation forest in the plurality of isolation forests comprises a plurality of isolation trees.
13. The method of claim 10 , wherein an isolation forest in the plurality of isolation forests without identified anomalies is labeled SAFE to indicate safe driving and the isolation forest with identified anomalies is labeled UNSAFE to indicate unsafe driving by using the vehicle telematics device.
14. The method of claim 10 , wherein calculating the driver risk score further comprises calculating the driver risk score by using an ensemble scoring process and the vehicle telematics device.
15. The method of claim 14 , wherein calculating the driver risks core by using an ensemble scoring process is evaluated by the vehicle telematics device by using the following expression:
Risk
Score
(
v
i
)
=
100
×
Num_UNUSAFE
n
where v i is a vehicle with a specific set of driver data, n is a total number of isolation forests in the plurality of isolation forests, and Num_UNSAFE is the number of isolation forests in the plurality of isolation forests labeled UNSAFE.
16. The method of claim 15 , wherein the driver risk score above a predetermined value indicates an unsafe driver.
17. The method of claim 10 , further comprising calculating the driver risk score for a specific date range by using the vehicle telematics device.
18. The method of claim 10 , wherein the communications device is a wireless device.