Device failure prediction using filter-based feature selection and a conformal prediction framework
An apparatus comprises a processing device configured to obtain monitoring data for devices comprising a plurality of features, determining two or more rankings of the features using two or more filter-based feature selection algorithms, and selecting a subset of the features based at least in part on the two or more rankings. The processing device is also configured to generate a failure prediction for a given one of the devices using at least one classifier that takes as input the selected subset of features, and applying a conformal prediction framework to the generated failure prediction to obtain a confidence measure indicating a quality of the generated failure prediction and a credibility measure indicating a quality of the monitoring data. The processing device is further configured to initiate one or more remedial actions based at least in part on the generated failure prediction, the confidence measure, and the credibility measure.
1. An apparatus comprising:
at least one processing device comprising a processor coupled to a memory;
the at least one processing device being configured to implement an analytics platform responsible for managing a plurality of devices in an information technology infrastructure, the analytics platform being configured to perform steps of:
obtaining monitoring data for the plurality of devices in the information technology infrastructure, the monitoring data comprising a plurality of features;
determining two or more rankings of the plurality of features using two or more filter-based feature selection algorithms;
generating an overall feature ranking for the plurality of features using a multi-criteria decision analysis method that integrates the two or more ranking of the plurality of features;
selecting a subset of the plurality of features based at least in part on the overall feature ranking for the plurality of features;
generating, utilizing at least a first machine learning algorithm in a machine learning system implemented by the processor and the memory of the at least one processing device, a failure prediction for each of the plurality of devices, wherein each of the failure predictions is generated using at least one classifier that takes as input the selected subset of the plurality of features, and wherein each of the failure predictions comprises a forecast value characterizing likelihood of one or more of the plurality of devices failing within a designated time frame;
applying a conformal prediction framework to the generated failure predictions, utilizing at least a second machine learning algorithm in the machine learning system implemented by the processor and the memory of the at least one processing device, to obtain confidence measures indicating a quality of each of the generated failure predictions and credibility measures indicating a quality of the monitoring data obtained for each of the plurality of devices;
selecting at least a given one of the plurality of devices for remediation based at least in part on (i) a given forecast value of a given one of the generated failure predictions for the given device; (ii) a given one of the confidence measures indicating the quality of the given generated failure prediction for the given device; and (iii) a given one of the credibility measures indicating the quality of the monitoring data obtained for the given device; and
prior to failure of the given device, in order to reduce at least one of data unavailability and data loss associated with failure of the given device, proactively migrating one or more workloads from the given device to one or more other ones of the plurality of devices in the information technology infrastructure and applying one or more fixes to one or more hardware components of the given device predicted to fail within the designated time frame.
2. The apparatus of claim 1 wherein the two or more filter-based feature selection algorithms comprise two or more of: a two sample t-test (T-TEST) algorithm; a Wilcoxon rank-sum (WRS) algorithm; a max-relevance min-redundancy (MRMR) algorithm; a mutual information maximization (MIM) algorithm; a mutual information feature selection (MIFS) algorithm; a joint mutual information (JMI) algorithm; a double input symmetrical relevance (DISR) algorithm; a conditional mutual information maximization (CMIM) algorithm; an interaction capping (ICAP) algorithm; and a conditional infomax feature extraction (CIFE) algorithm.
3. The apparatus of claim 1 wherein the multi-criteria decision analysis algorithm comprises a Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) multi-criteria decision analysis algorithm.
4. The apparatus of claim 1 wherein integrating the two or more rankings of the plurality of features utilizing the multi-criteria decision analysis algorithm comprises:
performing vector normalization for each of the two or more rankings of the plurality of features;
calculating, for each of the plurality of features across the two or more rankings of the plurality of features, distances from an ideal best ranking value and an ideal worst ranking value; and
calculating, for each of the plurality of features across the two or more rankings of the plurality of features, a performance score based at least in part on the distances from the ideal best ranking value and the ideal worst ranking value.
5. The apparatus of claim 1 wherein selecting the subset of the plurality of features comprises calculating performance scores for each of the plurality of features across the two or more ranking of the plurality of features and selecting a designated number of the plurality of features having the highest calculated performance scores.
6. The apparatus of claim 1 wherein the at least one classifier comprises at least one of: a support vector machine (SVM) classifier; a random forest (RF) classifier; a decision tree (DT) classifier; a linear discriminant analysis (LDA) classifier; a Naïve Bayes (NB) classifier; a fuzzy logic (FL) classifier; and a k nearest neighbor (KNN) classifier.
7. The apparatus of claim 1 wherein applying the conformal prediction framework comprises utilizing inductive conformal prediction (ICP) with an incremental stochastic gradient descent (SGD) machine learning algorithm.
8. The apparatus of claim 1 wherein the given device comprises a storage device.
9. The apparatus of claim 8 wherein:
migrating the one or more workloads from the given device to the one or more other ones of the plurality of devices in the information technology infrastructure comprises backing up data from the storage device to another storage device in the information technology infrastructure; and
applying the one or more fixes to the one or more hardware components of the given device predicted to fail within the designated time frame comprises at least one of:
applying a file system check to the storage device; and
replacing the storage device with another storage device.
10. The apparatus of claim 1 wherein the given device comprises a server.
11. The apparatus of claim 10 wherein:
migrating the one or more workloads from the given device to the one or more other ones of the plurality of devices in the information technology infrastructure comprises migrating one or more applications hosted on the server to another server in the information technology infrastructure; and
applying the one or more fixes to the one or more hardware components of the given device predicted to fail within the designated time frame comprises replacing at least one of one or more hardware components of the server.
12. A computer program product comprising a non-transitory processor-readable storage medium having stored therein program code of one or more software programs, wherein the program code when executed by at least one processing device implementing an analytics platform responsible for managing a plurality of devices in an information technology infrastructure, causes the at least one processing device to perform steps of:
obtaining monitoring data for the plurality of devices in the information technology infrastructure, the monitoring data comprising a plurality of features;
determining two or more rankings of the plurality of features using two or more filter-based feature selection algorithms;
generating an overall feature ranking for the plurality of features using a multi-criteria decision analysis method that integrates the two or more ranking of the plurality of features;
selecting a subset of the plurality of features based at least in part on the overall feature ranking for the plurality of features;
generating, utilizing at least a first machine learning algorithm in a machine learning system implemented by the at least one processing device, a failure prediction for each of the plurality of devices, wherein each of the failure predictions is generated using at least one classifier that takes as input the selected subset of the plurality of features, and wherein each of the failure predictions comprises a forecast value characterizing likelihood of one or more of the plurality of devices failing within a designated time frame;
applying a conformal prediction framework to the generated failure predictions, utilizing at least a second machine learning algorithm in the machine learning system implemented by the at least one processing device, to obtain confidence measures indicating a quality of each of the generated failure predictions and credibility measures indicating a quality of the monitoring data obtained for each of the plurality of devices;
selecting at least a given one of the plurality of devices for remediation based at least in part on (i) a given forecast value of a given one of the generated failure predictions for the given device; (ii) a given one of the confidence measures indicating the quality of the given generated failure prediction for the given device; and (iii) a given one of the credibility measures indicating the quality of the monitoring data obtained for the given device; and
prior to failure of the given device, in order to reduce at least one of data unavailability and data loss associated with failure of the given device, at proactively migrating one or more workloads from the given device to one or more other ones of the plurality of devices in the information technology infrastructure and applying one or more fixes to one or more hardware components of the given device predicted to fail within the designated time frame.
13. The computer program product of claim 12 wherein the multi-criteria decision analysis algorithm comprises a Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) multi-criteria decision analysis algorithm.
14. The computer program product of claim 12 wherein applying the conformal prediction framework comprises utilizing inductive conformal prediction (ICP) with an incremental stochastic gradient descent (SGD) machine learning algorithm.
15. The computer program product of claim 12 wherein the given device comprises a storage device, and wherein:
migrating the one or more workloads from the given device to the one or more other ones of the plurality of devices in the information technology infrastructure comprises backing up data from the storage device to another storage device in the information technology infrastructure; and
applying the one or more fixes to the one or more hardware components of the given device predicted to fail within the designated time frame comprises at least one of:
applying a file system check to the storage device; and
replacing the storage device with another storage device.
16. A method performed by an analytics platform responsible for managing a plurality of devices in an information technology infrastructure, the method comprising:
obtaining monitoring data for the plurality of devices in the information technology infrastructure, the monitoring data comprising a plurality of features;
determining two or more rankings of the plurality of features using two or more filter-based feature selection algorithms;
generating an overall feature ranking for the plurality of features using a multi-criteria decision analysis method that integrates the two or more ranking of the plurality of features;
selecting a subset of the plurality of features based at least in part on the overall feature ranking for the plurality of features;
generating, utilizing at least a first machine learning algorithm in a machine learning system, a failure prediction for each of the plurality of devices, wherein each of the failure predictions is generated using at least one classifier that takes as input the selected subset of the plurality of features, and wherein each of the failure predictions comprises a forecast value characterizing likelihood of one or more of the plurality of devices failing within a designated time frame;
applying a conformal prediction framework to the generated failure predictions, utilizing at least a second machine learning algorithm in the machine learning system, to obtain confidence measures indicating a quality of each of the generated failure predictions and credibility measures indicating a quality of the monitoring data obtained for each of the plurality of devices;
selecting at least a given one of the plurality of devices for remediation based at least in part on (i) a given forecast value of a given one of the generated failure predictions for the given device; (ii) a given one of the confidence measures indicating the quality of the given generated failure prediction for the given device; and (iii) a given one of the credibility measures indicating the quality of the monitoring data obtained for the given device; and
prior to failure of the given device, in order to reduce at least one of data unavailability and data loss associated with failure of the given device, proactively migrating one or more workloads from the given device to one or more other ones of the plurality of devices in the information technology infrastructure and applying one or more fixes to one or more hardware components of the given device predicted to fail within the designated time frame;
wherein the method is performed by at least one processing device comprising a processor coupled to a memory, the at least one processing device implementing the analytics platform and the machine learning system.
17. The method of claim 16 wherein the multi-criteria decision analysis algorithm comprises a Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) multi-criteria decision analysis algorithm.
18. The method of claim 16 wherein applying the conformal prediction framework comprises utilizing inductive conformal prediction (ICP) with an incremental stochastic gradient descent (SGD) machine learning algorithm.
19. The method of claim 16 wherein the given device comprises a storage device, and wherein:
migrating the one or more workloads from the given device to the one or more other ones of the plurality of devices in the information technology infrastructure comprises backing up data from the storage device to another storage device in the information technology infrastructure; and
applying the one or more fixes to the one or more hardware components of the given device predicted to fail within the designated time frame comprises at least one of:
applying a file system check to the storage device; and
replacing the storage device with another storage device.