Anomaly classifier

1. A method of classifying an anomaly, the method comprising:

receiving, from an anomaly detection system, one or more sets of time-series data as a data stream, each of the one or more sets of time-series data obtained from a respective sensor monitoring a machine, at least one of the one or more sets of time-series data comprising one or more anomalies;

a classifier transforming the anomalous time-series data using wavelets and detecting deviations in the time-series data;

grouping the time-series data into a plurality of groups including a time-series representation of the anomalous time-series data in their respective scale range, each range processed individually;

for each group of the plurality of groups, extracting, via a processor, statistical features including the deviations from the time-series data of each range;

creating a model for at least one of the one or more anomalies by combining extracted statistical features associated with the plurality of groups, the combining reducing dimensions provided to the classifier;

for each of the one or more anomalies, examining a respective closeness indicator, each respective closeness indicator specifying for a respective one of each of the one or more anomalies a respective temporal distance indicating how close or far away in time the respective anomaly is to a temporal location of normal data in at least one of the one or more sets of received time-series data;

based on the magnitude of each respective temporal distance, the classifier classifying each respective one of the one or more anomalies as one of normal data or not normal data based on the model of combined extracted statistical features; and

alerting a user to the anomaly classification of at least one of the one or more anomalies associated with the machine.

2. The method of claim 1 , wherein the scale range is related to brief oscillations associated with the received time-series data.

3. The method of claim 1 , wherein the plurality of groups comprises three or more scale ranges.

4. The method of claim 3 , wherein each of the plurality of groups is a time-series representation in their respective scale range.

5. The method of claim 1 , further comprising:

receiving a closeness indicator from the anomaly detection system; and

confirming, based on the closeness indicator, that the received time-series data is sufficiently far enough away from time-series data indicated as being normal to indicate an anomaly.

6. The method of claim 1 , wherein the plurality of groups is based on determining a scale range that defines a number of the groups in the plurality of groups.

7. A non-transitory computer-readable medium comprising instructions that when executed by a processor cause the processor to perform a method of classifying an anomaly, the method comprising:

a classifier transforming the anomalous time-series data using wavelets and detecting deviations in the time-series data;

grouping the time-series data into a plurality of groups including a time-series representation of the anomalous time-series data in their respective scale range, each range processed individually;

for each group of the plurality of groups, extracting, via a processor, statistical features including the deviations from the time-series data of each range;

creating a model for at least one of the one or more anomalies by combining extracted statistical features associated with the plurality of groups, the combining reducing dimensions provided to the classifier;

for each of the one or more anomalies, examining a respective closeness indicator, each respective closeness indicator specifying for a respective one of each of the one or more anomalies a respective temporal distance indicating how close or far away in time the respective anomaly is to a temporal location of normal data in at least one of the one or more sets of received time-series data;

based on the magnitude of each respective temporal distance, the classifier classifying each respective one of the one or more anomalies as one of normal data or not normal data based on the model of combined extracted statistical features; and

alerting a user to the anomaly classification of at least one of the one or more anomalies associated with the machine.

8. The medium of claim 7 , wherein the scale range is related to brief oscillations associated with the received time-series data.

9. The medium of claim 7 , wherein the plurality of groups comprises three or more scale ranges.

10. The medium of claim 9 , wherein each of the plurality of groups is a time-series representation in their respective scale range.

11. The medium of claim 7 , further comprising:

receiving a closeness indicator from the anomaly detection system; and

confirming, based on the closeness indicator, that the received time-series data is sufficiently far enough away from time-series data indicated as being normal to indicate an anomaly.

12. The method of claim 11 , wherein the closeness indicator is provided in terms of a percentage distance from the time-series data indicated as being normal.

13. The medium of claim 7 , wherein the plurality of groups is based on determining a scale range that defines a number of the groups in the plurality of groups.

14. A system for classifying an anomaly, the system comprising:

a processor; and

a non-transitory computer-readable medium comprising instructions that when executed by the processor cause the processor to perform a method to automatically detect anomalies, the method comprising:

monitoring a machine, at least one of the one or more sets of time-series data comprising one or more anomalies;

a classifier transforming the anomalous time-series data using wavelets and detecting deviations in the time-series data;

grouping the time-series data into a plurality of groups including a time-series representation of the anomalous time-series data in their respective scale range, each range processed individually;

for each group of the plurality of groups, extracting, via a processor, statistical features including the deviations from the time-series data of each range;

creating a model for at least one of the one or more anomalies by combining extracted statistical features associated with the plurality of groups, the combining reducing dimensions provided to the classifier;

for each of the one or more anomalies, examining a respective closeness indicator, each respective closeness indicator specifying for a respective one of each of the one or more anomalies a respective temporal distance indicating how close or far away in time the respective anomaly is to a temporal location of normal data in at least one of the one or more sets of received time-series data;

based on the magnitude of each respective temporal distance, the classifier classifying each respective one of the one or more anomalies as one of normal data or not normal data based on the model of combined extracted statistical features; and

alerting a user to the anomaly classification of at least one of the one or more anomalies associated with the machine.

15. The system of claim 14 , wherein the scale range is related to brief oscillations associated with the received time-series data.

16. The system of claim 14 , wherein the plurality of groups comprises three or more scale ranges.

17. The system of claim 16 , wherein each of the plurality of groups is a time-series representation in their respective scale range.

18. The system of claim 14 , further comprising:

receiving a closeness indicator from the anomaly detection system; and

confirming, based on the closeness indicator, that the received time-series data is sufficiently far enough away from time-series data indicated as being normal to indicate an anomaly.

19. The system of claim 14 , wherein the plurality of groups is based on determining a scale range that defines a number of the groups in the plurality of groups.