Anomaly detection in a network based on a key performance indicator prediction model
A network monitoring platform may obtain a measurement of a particular value of a key performance indicator (KPI) and one or more parameters of the particular value of the KPI. The network monitoring platform may determine a prediction of the particular value of the KPI. The network monitoring platform may determine an amount of error in the prediction of the particular value of the KPI, wherein the amount of error in the prediction of the particular value of the KPI is based on a difference between the prediction of the particular value of the KPI and the measurement of the particular value of the KPI. The network monitoring platform may perform, based on the amount of error in the prediction of the particular value of the KPI, one or more actions.
1. A method, comprising:
determining, by a device, a prediction of a particular value of a key performance indicator (KPI) based on a KPI prediction model trained on historical values of the KPI;
determining, by the device, an amount of error in the prediction of the particular value of the KPI,
wherein the amount of error in the prediction of the particular value of the KPI is based on a difference between the prediction of the particular value of the KPI and a measurement of the particular value of the KPI;
determining, by the device, a weighting for a weighted average of the amount of error in the prediction of the particular value of the KPI and amounts of respective errors in predictions of the historical values of the KPI,
wherein determining the weighting comprises:
determining the weighting based on respective ages of the predictions of the historical values of the KPI; and
determining, by the device and based on determining the weighting, a prediction accuracy value based on the amount of error in the prediction of the particular value of the KPI.
2. The method of claim 1 , wherein the KPI prediction model is trained using one or more machine learning processes.
3. The method of claim 1 , wherein the historical values of the KPI are measured during a particular period of time.
4. The method of claim 1 , wherein determining the weighting comprises:
determining the weighting based on a subset of the amounts of respective errors in the predictions of the historical values of the KPI,
wherein the subset includes the amounts of respective errors in the predictions of the historical values of the KPI within a particular time threshold.
5. The method of claim 1 , further comprising:
training the KPI prediction model using a neural network.
6. The method of claim 1 , wherein determining the amount of error in the prediction of the particular value of the KPI comprises:
determining whether the amount of error in the prediction of the particular value of the KPI satisfies an anomaly threshold; and
further comprising:
updating the anomaly threshold based on determining the prediction accuracy value.
7. A device, comprising:
one or more memories; and
one or more processors, coupled to the one or more memories, configured to:
determine a prediction of a particular value of a key performance indicator (KPI) based on a KPI prediction model trained on historical values of the KPI;
determine an amount of error in the prediction of the particular value of the KPI,
wherein the amount of error in the prediction of the particular value of the KPI is based on a difference between the prediction of the particular value of the KPI and a measurement of the particular value of the KPI;
determine a weighting for a weighted average of the amount of error in the prediction of the particular value of the KPI and amounts of respective errors in predictions of the historical values of the KPI,
wherein the one or more processors, when determining the weighting, are configured to:
determine the weighting based on respective ages of the predictions of the historical values of the KPI; and
determine, based on determining the weighting, a prediction accuracy value based on the amount of error in the prediction of the particular value of the KPI.
8. The device of claim 7 , wherein the KPI prediction model is trained using one or more machine learning processes.
9. The device of claim 7 , wherein the historical values of the KPI are measured during a particular period of time.
10. The device of claim 7 , wherein the one or more processors, when determining the weighting, are configured to:
determine the weighting based on a subset of the amounts of respective errors in the predictions of the historical values of the KPI,
wherein the subset includes the amounts of respective errors in the predictions of the historical values of the KPI within a particular time threshold.
11. The device of claim 7 , wherein the one or more processors are further configured to:
train the KPI prediction model using a neural network.
12. The device of claim 7 , wherein the one or more processors, when determining the amount of error in the prediction of the particular value of the KPI, are configured to:
determine whether the amount of error in the prediction of the particular value of the KPI satisfies an anomaly threshold; and
wherein the one or more processors are further configured to:
update the anomaly threshold based on determining the prediction accuracy value.
13. 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:
determine a prediction of a particular value of a key performance indicator (KPI) based on a KPI prediction model trained on historical values of the KPI;
determine an amount of error in the prediction of the particular value of the KPI,
wherein the amount of error in the prediction of the particular value of the KPI is based on a difference between the prediction of the particular value of the KPI and a measurement of the particular value of the KPI;
determine a weighting for a weighted average of the amount of error in the prediction of the particular value of the KPI and amounts of respective errors in predictions of the historical values of the KPI,
wherein the one or more instructions, that cause the device to determine the weighting, cause the device to:
determine the weighting based on respective ages of the predictions of the historical values of the KPI; and
determine, based on determining the weighting, a prediction accuracy value based on the amount of error in the prediction of the particular value of the KPI.
14. The non-transitory computer-readable medium of claim 13 , wherein the KPI prediction model is trained using one or more machine learning processes.
15. The non-transitory computer-readable medium of claim 13 , wherein the historical values of the KPI are measured during a particular period of time.
16. The non-transitory computer-readable medium of claim 13 , wherein the one or more instructions, that cause the device to determine the weighting, cause the device to:
determine the weighting based on a subset of the amounts of respective errors in the predictions of the historical values of the KPI,
wherein the subset includes the amounts of respective errors in the predictions of the historical values of the KPI within a particular time threshold.
17. The non-transitory computer-readable medium of claim 13 , wherein the one or more instructions further cause the device to:
train the KPI prediction model using a neural network.
18. The method of claim 1 , further comprising:
determining the particular value of the KPI is anomalous based on the prediction accuracy value.
19. The device of claim 7 , wherein the one or more processors are further configured to:
determine a level of severity associated with the particular value of the KPI based on the prediction accuracy value.
20. The non-transitory computer-readable medium of claim 13 , wherein the one or more instructions further cause the device to:
retrain the KPI prediction model based on the particular value of the KPI.