Systems and methods for using machine learning for managing application incidents
Disclosed herein are systems and methods for using machine learning for managing application incidents. An embodiment takes the form of a method that includes receiving extracted data pertaining to one or more applications, Model-input data is generated from the extracted data. Model-output data is generated at least in part by processing the generated model-input data with one or more machine-learning models trained to make one or more application-incident predictions. Based at least in part on the model-output data, an application-incident-likely determination is made that a likelihood of an occurrence of an application incident exceeds an application-incident-likelihood threshold, where the application incident corresponds to a given application of the one or more applications. Responsive to making the application-incident-likely determination, one or more alerts of the likelihood of the occurrence of the application incident are output.
1. A method comprising:
receiving extracted data pertaining to one or more applications;
generating model-input data from the extracted data;
generating model-output data at least in part by processing the generated model-input data with a plurality of machine-learning models each independently trained to make one or more application-incident predictions, wherein the plurality of machine-learning models comprises a plurality of incident-type-specific machine-learning models comprising:
a first machine-learning model that is trained to make application-incident predictions with respect to a first type of application incident; and
a second machine-learning model that is trained to make application-incident predictions with respect to a second type of application incident,
the first type of application incident and the second type of application incident being different from one another;
making, based at least in part on the model-output data, an application-incident-likely determination that a likelihood of an occurrence of an application incident exceeds an application-incident-likelihood threshold, the application incident corresponding to a given application of the one or more applications; and
responsive to making the application-incident-likely determination, outputting one or more alerts of the likelihood of the occurrence of the application incident.
2. The method of claim 1 , wherein:
the model-input data is structured according to a set of one or more features;
generating the model-input data from the extracted data comprises transforming at least a portion of the extracted data into being structured according to the set of one or more features; and
processing the generated model-input data with the one or more machine-learning models comprises processing the set of one or more features with the one or more machine-learning models.
3. The method of claim 2 , wherein:
receiving the extracted data comprises receiving the extracted data as a plurality of datasets respectively extracted from a plurality of different data stores; and
transforming at least a portion of the extracted data into being structured according to the set of one or more features comprises one or more of:
conducting at least one normalization function with respect to the at east a portion of the extracted data;
conducting at east one join operation with respect to the at least a portion of the extracted data;
conducting at east one metric calculation with respect to the at least a portion of the extracted data; and
conducting at least one data-quality check with respect to the at least a portion of the extracted data.
4. The method of claim 1 , wherein:
a data-shaping platform generates the model-input data from the extracted data;
a machine-learning platform generates the model-output data at least in part by processing the generated model-input data with the plurality of machine-learning models; and
the method further comprises conveying the model-input data from the data-shaping platform to the machine-learning platform using data-movement software.
5. The method of claim 1 , wherein the plurality of machine-learning models comprises one or more gradient boosting machine (GBM) models.
6. The method of claim 1 , wherein the first type of application incident comprises an application-patching-related incident.
7. The method of claim 1 , wherein the first type of application incident comprises an application-access-related incident.
8. The method of claim 1 , wherein the first type of application incident comprises an application-configuration-related incident.
9. The method of claim 1 , wherein the first type of application incident comprises an application-server-relationship-related incident.
10. The method of claim 1 , wherein the plurality of machine-learning models further comprises:
a third machine-learning model that is trained to make application-incident predictions with respect to a first application of the one or more applications; and
a fourth machine-learning model that is trained to make application-incident predictions with respect to a second application of the one or more applications,
the first application and the second application being different from one another.
11. The method of claim 1 , wherein the application-incident-likely determination is that the likelihood of an occurrence of an application incident within a predetermined amount of time exceeds the application-incident-likelihood threshold.
12. The method of claim 1 , wherein:
the model-output data indicates the likelihood of the occurrence of the application incident; and
making the application-incident-likely determination based at least in part on the model-output data comprises comparing the indicated likelihood to the application-incident-likelihood threshold.
13. The method of claim 1 , wherein:
the model-output data comprises an indication that the likelihood of the occurrence of the application incident exceeds the application-incident-likelihood threshold; and
making the application-incident-likely determination based at least in part on the model-output data comprises making the application-incident-likely determination based at least in part on the indication.
14. The method of claim 1 , further comprising presenting the one or more alerts via one or more user interfaces.
15. The method of claim 1 , wherein outputting one or more alerts comprises outputting the one or more alerts to one or more of data storage, a computing device, and a networked server.
16. The method of claim 1 , further comprising outputting the model-output data to an administrative interface.
17. A system comprising:
at least one processor; and
one or more non-transitory computer readable storage media containing instructions executable by the at least one processor for causing the at least one processor to perform operations comprising:
receiving extracted data pertaining to one or more applications;
generating model-input data from the extracted data;
generating model-output data at least in part by processing the generated model-input data with a plurality of machine-learning models each independently trained to make one or more application-incident predictions wherein the plurality of machine-learning models comprises a plurality of incident-type-specific machine-learning models comprising:
a first machine-learning model that is trained to make application-incident predictions with respect to a first type of application incident; and
a second machine-learning model that is trained to make application-incident predictions with respect to a second type of application incident,
the first type of application incident and the second type of application incident being different from one another;
making, based at least in part on the model-output data, an application-incident-likely determination that a likelihood of an occurrence of an application incident exceeds an application-incident-likelihood threshold, the application incident corresponding to a given application of the one or more applications; and
responsive to making the application-incident-likely determination, outputting one or more alerts of the likelihood of the occurrence of the application incident.
18. One or more non-transitory computer readable storage media containing instructions executable by at least one processor for causing the at least one processor to perform operations comprising:
receiving extracted data pertaining to one or more applications;
generating model-input data from the extracted data;
generating model-output data at least in part by processing the generated model-input data with a plurality of machine-learning models each independently trained to make one or more application-incident predictions, wherein the plurality of machine-learning models comprises a plurality of incident-type-specific machine-learning models comprising:
a first machine-learning model that is trained to make application-incident predictions with respect to a first type of application incident; and
a second machine-learning model that is trained to make application-incident predictions with respect to a second type of application incident,
the first type of application incident and the second type of application incident being different from one another:
making, based at least in part on the model-output data, an application-incident-likely determination that a likelihood of an occurrence of an application incident exceeds an application-incident-likelihood threshold, the application incident corresponding to a given application of the one or more applications; and
responsive to making the application-incident-likely determination, outputting one or more alerts of the likelihood of the occurrence of the application incident.