Anomaly prediction for electronic resources
Historical time-series data can be analyzed using a probabilistic model to determine one or more distributions, including at least a normal distribution and an anomaly distribution. These distributions can be analyzed to obtain values for distribution parameters, such as mean, standard deviation, and density, as well as other statistical parameters, for use in building a forecasting model. This model can analyze the time-series data to predict or forecast actionable anomalies at one or more future points or periods in time, such as may exceed a determined anomaly threshold with at least a minimum amount of confidence. A determination can be made as to one or more actions to take in anticipation of the anomalous event, or volume of events, such as to attempt to prevent the occurrence or to be better positioned to handle the occurrence. Such forecasting or prediction can utilize both modeling and feature engineering.
1. A computer-implemented method, comprising:
obtaining historical volume data representative of a plurality of past events;
generating, from the historical volume data, a set of time-series data;
analyzing the time-series data using a Gaussian mixture model to determine a normal distribution and an anomaly distribution;
determining, from the normal distribution and the anomaly distribution, values for a set of distribution parameters;
generating, using the values for the set of distribution parameters, a time-series forecasting model that comprises a machine learning model trained using probabilities for different anomalous events;
determining, using the time-series forecasting model, a quantity of future anomalies; and
providing notification of the quantity, wherein at least one proactive action is enabled to be taken for an occurrence of the future anomalies.
2. The computer-implemented method of claim 1 , further comprising:
determining whether the quantity of future anomalies exceeds an anomaly threshold, with at least a minimum confidence, before taking the at least one proactive action.
3. The computer-implemented method of claim 1 ,
wherein the quantity of future anomalies is further determined using a forecasting framework that provides for feature engineering, along with the time-series forecasting model, to be associated with event forecasting.
4. The computer-implemented method of claim 3 , wherein the machine learning model is trained using Gaussian prediction model parameters and a quantity of an anomalous event predicted for at least one future time or date.
5. The computer-implemented method of claim 1 , wherein the set of distribution parameters includes at least mean, standard deviation, and density.
6. A computer-implemented method, comprising:
determining a normal distribution and an anomaly distribution from historical time-series data;
determining, for the normal distribution and the anomaly distribution, values for a set of distribution parameters;
building a time-series forecasting model based upon the values for the set of distribution parameters, the time-series forecasting model comprising a machine learning model trained using probabilities for different anomalous events; and
generating, using the time-series forecasting model, a prediction of a future occurrence of one or more anomalies.
7. The computer-implemented method of claim 6 , wherein the normal distribution and the anomaly distribution are determined using a Gaussian mixture model.
8. The computer-implemented method of claim 6 , further comprising:
determining at least one action to take based upon the prediction, the at least one action selected to manage, prevent, or lessen an impact of the future occurrence of the one or more anomalies.
9. The computer-implemented method of claim 6 , wherein an additional distribution is determined from the historical time series data, and wherein the anomaly distribution is determined from the normal distribution and the additional distribution, the additional distribution including a Gaussian distribution or a binomial distribution.
10. The computer-implemented method of claim 6 , further comprising:
determining a probability threshold, wherein a volume of events of a given type that exceeds the probability threshold at a future point or period in time is classified as an anomaly.
11. The computer-implemented method of claim 6 , further comprising:
classifying, using the machine learning model, the one or more anomalies as actionable anomalies if the one or more anomalies are predicted with at least a minimum confidence or probability.
12. The computer-implemented method of claim 6 , wherein the prediction of the future occurrence is further determined using a framework that provides for feature engineering along with the time-series forecasting model.
13. The computer-implemented method of claim 12 , wherein the machine learning model is trained using Gaussian prediction model parameters and a quantity of an anomalous event predicted for at least one future time or date.
14. The computer-implemented method of claim 6 , wherein the set of distribution parameters includes at least mean, standard deviation, and density.
15. The computer-implemented method of claim 6 , further comprising:
generating the historical time-series data from historical volume data.
16. A system, comprising:
a processor; and
memory including instructions that, when executed by the processor, cause the system to:
determine an event distribution and an anomaly distribution from historical time-series data;
determine, for the event distribution and the anomaly distribution, values for a set of distribution parameters;
obtain a time-series forecasting model based upon the values for the set of distribution parameters, the time-series forecasting model comprising a machine learning model trained using probabilities for different anomalous events; and
generate, using the time-series forecasting model, a prediction of a future occurrence of one or more anomalies.
17. The system of claim 16 , wherein the event distribution and the anomaly distribution are determined using a Gaussian mixture model.
18. The system of claim 16 , wherein the instructions when executed further cause the system to:
determine at least one action to take based upon the prediction, the at least one action selected to manage, prevent, or lessen an impact of the future occurrence of the one or more anomalies.
19. The system of claim 16 , wherein the instructions when executed further cause the system to:
determine an anomaly threshold, wherein a volume of events of a given type that exceeds the anomaly threshold at a future point or period in time is classified as an anomaly.
20. The system of claim 16 , wherein the prediction of the future occurrence is further determined using a framework that provides for feature engineering along with the time-series forecasting model, and wherein the machine learning model is trained using Gaussian prediction model parameters and a quantity of an anomalous event predicted for at least one future time or date.