Systems and methods for determining when to relabel data for a machine learning model

1. A method, comprising:

receiving, by a device, video data identifying videos associated with driving events of vehicles;

processing, by the device, the video data, with a machine learning model, to determine classifications for the videos;

generating, by the device, labels for the videos based on the classifications;

calculating, by the device, event severity scores and event severity labels based on the classifications and the labels;

calculating, by the device, event severity incoherence scores based on the event severity scores and the event severity labels;

calculating, by the device, user feedback scores based on feedback votes and suggested event severities provided by users associated with the device;

determining, by the device, reviewer mistrust scores based on a quantity of incorrect reviews and a quantity of all reviews provided by reviewers;

calculating, by the device, time review scores based on review time distributions associated with reviews provided by the reviewers;

calculating, by the device, reviewer bias scores based on reviewer label bias and a quantity of the labels;

determining, by the device, relabeling scores for the videos based on the event severity incoherence scores, the user feedback scores, the reviewer mistrust scores, the time review scores, and the reviewer bias scores;

generating, by the device, one or more new labels for one or more of the videos based on the relabeling scores for the videos; and

retraining, by the device, the machine learning model, with the one or more new labels, to generate a retrained machine learning model.

2. The method of claim 1 , further comprising:

storing the one or more new labels in a data structure that includes one or more labels, of the labels for the videos, that are not replaced with the one or more new labels.

3. The method of claim 1 , further comprising:

implementing the retrained machine learning model with new video data identifying new videos associated with new driving events of the vehicles.

4. The method of claim 1 , wherein calculating the event severity scores and the event severity labels based on the classifications and the labels comprises:

calculating the event severity scores based on applying weights to the classifications; and

calculating the event severity labels based on applying values to the labels.

5. The method of claim 1 , wherein calculating the event severity incoherence scores based on the event severity scores and the event severity labels comprises:

calculating the event severity incoherence scores based on applying a distance measure to the event severity scores and the event severity labels.

6. The method of claim 1 , wherein calculating the user feedback scores based on the feedback votes and the suggested event severities comprises:

subtracting the suggested event severities from the event severity labels to obtain first values;

dividing the first values by second values to obtain third values;

subtracting the feedback votes from fourth values to obtain fifth values;

dividing the fifth values by sixth values to obtain seventh values; and

multiplying the third values and the seventh values to calculate the user feedback score.

7. The method of claim 1 , wherein determining the reviewer mistrust scores based on the quantity of incorrect reviews and the quantity of all reviews comprises:

dividing the quantity of all reviews by first values to obtain second values; and

dividing the quantity of incorrect reviews by the second values to determine the reviewer mistrust scores.

8. A device, comprising:

one or more processors configured to:

receive video data identifying videos associated with driving events of vehicles;

process the video data, with a machine learning model, to determine classifications for the videos;

generate labels for the videos based on the classifications;

calculate event severity scores and event severity labels based on the classifications and the labels;

calculate event severity incoherence scores based on the event severity scores and the event severity labels;

calculate user feedback scores based on feedback votes and suggested event severities provided by users associated with the device;

determine reviewer mistrust scores based on a quantity of incorrect reviews and a quantity of all reviews provided by reviewers;

calculate time review scores based on review time distributions associated with reviews provided by the reviewers;

calculate reviewer bias scores based on reviewer label bias and a quantity of the labels;

determine relabeling scores for the videos based on the event severity incoherence scores, the user feedback scores, the reviewer mistrust scores, the time review scores, and the reviewer bias scores;

generate one or more new labels for one or more of the videos based on the relabeling scores for the videos;

retrain the machine learning model, with the one or more new labels, to generate a retrained machine learning model; and

implement the retrained machine learning model.

9. The device of claim 8 , wherein the one or more processors, to calculate the time review scores based on the review time distributions, are configured to:

determine model times elapsed to generate the reviews provided by the reviewers;

generate the review time distributions based on the model times; and

calculate the time review scores based on generating the review time distributions.

10. The device of claim 8 , wherein the one or more processors, to calculate the reviewer bias scores based on the reviewer label bias and the quantity of the labels, are configured to:

divide the reviewer label bias, for each of the reviewers, by the quantity of labels generated by each of the reviewers to generate a first value for each of the reviewers; and

add the first value for each of the reviewers to calculate the reviewer bias scores.

11. The device of claim 8 , wherein the one or more processors, to determine the relabeling scores for the videos, are configured to:

multiply the reviewer mistrust scores, the time review scores, and the reviewer bias scores to obtain first values; and

add the event severity incoherence scores, the user feedback scores, the first values, and second values to determine the relabeling scores for the videos.

12. The device of claim 8 , wherein the one or more processors are further configured to:

determine that a new label is generated for one of the videos more than a threshold quantity of times; and

discard the one of the videos based on determining that a new label is generated for one of the videos more than the threshold quantity of times.

13. The device of claim 8 , wherein the one or more processors are further configured to:

determine whether the relabeling scores satisfy a score threshold,

wherein the one or more processors, to generate the one or more new labels for the one or more of the videos, are configured to:

selectively:

generate a new label for one of the videos based on one of the relabeling scores satisfying the score threshold, or

not generate a new label for one of the videos based on one of the relabeling scores failing to satisfy the score threshold.

14. The device of claim 8 , wherein the one or more processors are further configured to:

determine whether the relabeling scores satisfy a score threshold,

wherein the one or more processors, to generate the one or more new labels for the one or more of the videos, are configured to:

generate the one or more new labels for the one or more of the videos based on one or more relabeling scores, associated with the one or more videos, satisfying the score threshold.

15. A non-transitory computer-readable medium storing a set of instructions, the set of instructions comprising:

one or more instructions that, when executed by one or more processors of a device, cause the device to:

receive video data identifying videos associated with driving events of vehicles;

process the video data, with a machine learning model, to determine classifications for the videos;

generate labels for the videos based on the classifications;

calculate event severity scores and event severity labels based on the classifications and the labels;

calculate event severity incoherence scores based on the event severity scores and the event severity labels;

calculate user feedback scores based on feedback votes and suggested event severities provided by users associated with the device;

determine reviewer mistrust scores based on a quantity of incorrect reviews and a quantity of all reviews provided by reviewers;

calculate time review scores based on review time distributions associated with reviews provided by the reviewers;

calculate reviewer bias scores based on reviewer label bias and a quantity of the labels;

determine relabeling scores for the videos based on the event severity incoherence scores, the user feedback scores, the reviewer mistrust scores, the time review scores, and the reviewer bias scores;

generate one or more new labels for one or more of the videos based on the relabeling scores for the videos;

store the one or more new labels in a data structure that includes the labels that are not replaced with the one or more new labels; and

retrain the machine learning model, with the one or more new labels, to generate a retrained machine learning model.

16. The non-transitory computer-readable medium of claim 15 , wherein the one or more instructions further cause the device to:

implement the retrained machine learning model with new video data identifying new videos associated with new driving events of the vehicles.

17. The non-transitory computer-readable medium of claim 15 , wherein the one or more instructions, that cause the device to calculate the event severity scores and the event severity labels based on the classifications and the labels, cause the device to:

calculate the event severity scores based on applying weights to the classifications; and

calculate the event severity labels based on applying values to the labels.

18. The non-transitory computer-readable medium of claim 15 , wherein the one or more instructions, that cause the device to calculate the event severity incoherence scores based on the event severity scores and the event severity labels, cause the device to:

calculate the event severity incoherence scores based on applying a distance measure to the event severity scores and the event severity labels.

19. The non-transitory computer-readable medium of claim 15 , wherein the one or more instructions, that cause the device to calculate the user feedback scores based on the feedback votes and the suggested event severities, cause the device to:

subtract the suggested event severities from the event severity labels to obtain first values;

divide the first values by second values to obtain third values;

subtract the feedback votes from fourth values to obtain fifth values;

divide the fifth values by sixth values to obtain seventh values; and

multiply the third values and the seventh values to calculate the user feedback score.

20. The non-transitory computer-readable medium of claim 15 , wherein the one or more instructions, that cause the device to determine the reviewer mistrust scores based on the quantity of incorrect reviews and the quantity of all reviews, cause the device to:

divide the quantity of all reviews by first values to obtain second values; and

divide the quantity of incorrect reviews by the second values to determine the reviewer mistrust scores.