Anomaly detection
Disclosed is a computing platform including a memory assembly having encoded thereon executable control-logic instructions configured to be executable by the computing platform, and also configured to urge the computing platform to carry out a method comprising receiving data; and detecting at least one anomaly contained in the data that was received.
1. A system for detecting vehicle data volume anomalies, the system comprising:
at least one datastore operable to store vehicle data originating from a plurality of monitoring devices installed in a plurality of vehicles;
at least one processor in communication with the at least one data store, the at least one processor operable to:
receive, during a training period, a first collection of vehicle data from the plurality of monitoring devices;
train a machine learning model using the first collection of vehicle data to forecast a future amount of vehicle data to be received from the plurality of monitoring devices;
determine, based on the machine learning model, a plurality of confidence intervals for the predicted future amount of vehicle data;
determine that no confidence interval satisfies a percentile width of the first collection of vehicle data;
iteratively retrain the machine learning model using a subset of the first collection of vehicle data until at least one of the confidence intervals satisfies the percentile width of the first collection of vehicle data;
select one of the confidence intervals that satisfies the percentile width of the first collection of vehicle data;
receive, during a forecasting period, a second collection of vehicle data from the plurality of monitoring devices;
detect a vehicle data volume anomaly in the second collection of vehicle data by determining that the amount of vehicle data in the second collection of vehicle data is outside of the selected confidence interval; and
generate a notification indicating the detection of the vehicle data volume anomaly.
2. The system of claim 1 , wherein selecting one of the confidence intervals comprises selecting a confidence interval that is narrower than the percentile width of the first collection of vehicle data.
3. The system of claim 1 , wherein the machine learning model is an Auto-Regressive Integrated Moving Average (ARIMA) model.
4. The system of claim 1 , wherein the percentile width is a distance between the twentieth and eightieth percentile of the amount of vehicle data in the second collection of vehicle data.
5. The system of claim 1 , wherein the plurality of confidence intervals comprises five percent increments between fifty percent to night-five percent.
6. The system of claim 1 , wherein the at least one processor is operable to:
determine a severity level for the vehicle data volume anomaly based on the amount of vehicle data in the second collection of vehicle data; and
determine a type of notification based on the severity level for the vehicle data volume anomaly.
7. The system of claim 1 , wherein the at least one processor is operable to:
determine whether the second collection of vehicle data corresponds to a holiday; and
in response to determining that the second collection of vehicle data corresponds to a holiday, stop the notification indicating the detection of the vehicle data volume anomaly.
8. The system of claim 1 , wherein the at least one processor is operable to:
determine a high volume anomaly when the amount of vehicle data in the second collection of vehicle data is above an upper bound of the selected confidence interval; and
determine a low volume anomaly when the amount of vehicle data in the second collection of vehicle data is below a lower bound of the selected confidence interval.
9. The system of claim 1 , wherein the at least one processor is operable to:
receive, during a second training period, a third collection of vehicle data from the plurality of monitoring devices; and
retrain the machine learning model using the third collection of vehicle data to forecast another future amount of vehicle data to be received from the plurality of monitoring devices.
10. A method for detecting vehicle data volume anomalies, the method comprising operating at least one processor to:
receive, during a training period, a first collection of vehicle data from a plurality of monitoring devices installed in a plurality of vehicles;
train a machine learning model using the first collection of vehicle data to forecast a future amount of vehicle data to be received from the plurality of monitoring devices;
determine, based on the machine learning model, a plurality of confidence intervals for the predicted future amount of vehicle data;
determine that no confidence interval satisfies a percentile width of the first collection of vehicle data;
iteratively retrain the machine learning model using a subset of the first collection of vehicle data until at least one of the confidence intervals satisfies the percentile width of the first collection of vehicle data;
select one of the confidence intervals that satisfies the percentile width of the first collection of vehicle data;
receive, during a forecasting period, a second collection of vehicle data from the plurality of monitoring devices;
detect a vehicle data volume anomaly in the second collection of vehicle data by determining that the amount of vehicle data in the second collection of vehicle data is outside of the selected confidence interval; and
generate a notification indicating the detection of the vehicle data volume anomaly.
11. The method of claim 10 , wherein selecting one of the confidence intervals comprises selecting a confidence interval that is narrower than the percentile width of the first collection of vehicle data.
12. The method of claim 10 , wherein the machine learning model is an Auto-Regressive Integrated Moving Average (ARIMA) model.
13. The method of claim 10 , wherein the percentile width is a distance between the twentieth and eightieth percentile of the amount of vehicle data in the second collection of vehicle data.
14. The method of claim 10 , wherein the plurality of confidence intervals comprises five percent increments between fifty percent to night-five percent.
15. The method of claim 10 , further comprising operating the at least one processor to:
determine a severity level for the vehicle data volume anomaly based on the amount of vehicle data in the second collection of vehicle data; and
determine a type of notification based on the severity level for the vehicle data volume anomaly.
16. The method of claim 10 , further comprising operating the at least one processor to:
determine whether the second collection of vehicle data corresponds to a holiday; and
in response to determining that the second collection of vehicle data corresponds to a holiday, stop the notification indicating the detection of the vehicle data volume anomaly.
17. The method of claim 10 , further comprising operating the at least one processor to:
determine a high volume anomaly when the amount of vehicle data in the second collection of vehicle data is above an upper bound of the selected confidence interval; and
determine a low volume anomaly when the amount of vehicle data in the second collection of vehicle data is below a lower bound of the selected confidence interval.
18. The method of claim 10 , further comprising operating the at least one processor to:
receive, during a second training period, a third collection of vehicle data from the plurality of monitoring devices; and
retrain the machine learning model using the third collection of vehicle data to forecast another future amount of vehicle data to be received from the plurality of monitoring devices.