Machine logic for performing anomaly detection
Technology for computerized anomaly detection, where the machine logic (for example, software) utilizes a multi-layer anomaly detection method that may include three (3) layers: (a) a baseline model for each feature at a single point, (b) a dynamic time window for historical data, and/or (c) correlation analysis for different model features.
1. A computer implemented method (CIM) comprising:
receiving current input data including a stream of first feature values corresponding to a first measured parameter that is used to characterize some aspect of an ongoing production process used to produce a product;
creating a first baseline model for the first feature, with the baseline model indicating data patterns indicated a normal range and a set of anomaly-indicative data patterns that indicates potential existence of an anomaly occurring in the ongoing production process;
applying the first baseline model to first feature values being received in the first stream of data values to tentatively determine an anomaly in the ongoing production process;
responsive to the tentative determination of the anomaly, locating a set of historical time period(s) when the stream of first feature values followed a similar pattern to the current input data; and
making a final determination that an anomaly is not occurring based, at least in part, on historical first feature data values from the set of historical time period(s).
2. The CIM of claim 1 wherein the product is a foodstuff.
3. The CIM of claim 2 wherein the first feature is temperature.
4. The CIM of claim 1 further comprising:
using a combination of single timepoint based model and period based modelwherein the first feature is reflectivity.
5. The CIM of claim 1 wherein the first feature is a physical dimension of instantiations of the product being produced.
6. A computer program product (CPP) comprising:
a set of storage device(s); and
non-transitory computer code stored collectively in the set of storage device(s), with the computer code including data and instructions to cause a processor(s) set to perform at least the following operations:
receiving current input data including a stream of first feature values corresponding to a first measured parameter that is used to characterize some aspect of an ongoing production process used to produce a product,
creating a first baseline model for the first feature, with the baseline model indicating data patterns indicated a normal range and a set of anomaly-indicative data patterns that indicates potential existence of an anomaly occurring in the ongoing production process,
applying the first baseline model to first feature values being received in the first stream of data values to tentatively determine an anomaly in the ongoing production process,
responsive to the tentative determination of the anomaly, locating a set of historical time period(s) when the stream of first feature values followed a similar pattern to the current input data, and
making a final determination that an anomaly is not occurring based, at least in part, on historical first feature data values from the set of historical time period(s).
7. The CPP of claim 6 wherein the product is a foodstuff.
8. The CPP of claim 7 wherein the first feature is temperature.
9. The CPP of claim 6 wherein the first feature is reflectivity.
10. The CPP of claim 6 wherein the first feature is a physical dimension of instantiations of the product being produced.
11. A computer system (CS) comprising:
a processor(s) set;
a set of storage device(s); and
computer code stored collectively in the set of storage device(s), with the computer code including data and instructions to cause the processor(s) set to perform at least the following operations:
receiving current input data including a stream of first feature values corresponding to a first measured parameter that is used to characterize some aspect of an ongoing production process used to produce a product,
creating a first baseline model for the first feature, with the baseline model indicating data patterns indicated a normal range and a set of anomaly-indicative data patterns that indicates potential existence of an anomaly occurring in the ongoing production process,
applying the first baseline model to first feature values being received in the first stream of data values to tentatively determine an anomaly in the ongoing production process,
responsive to the tentative determination of the anomaly, locating a set of historical time period(s) when the stream of first feature values followed a similar pattern to the current input data, and
making a final determination that an anomaly is not occurring based, at least in part, on historical first feature data values from the set of historical time period(s).
12. The CS of claim 11 wherein the product is a foodstuff.
13. The CS of claim 12 wherein the first feature is temperature.
14. The CS of claim 11 wherein the first feature is reflectivity.
15. The CS of claim 11 wherein the first feature is a physical dimension of instantiations of the product being produced.