Determining failure modes of devices based on text analysis

• US 11,922,377 B2
• Filed: 03/20/2018
• Issued: 03/05/2024
• Est. Priority Date: 10/24/2017
• Status: Active Grant

• Current Assignee: SAP SE
• Original Assignee: SAP SE
• Inventors: Shetty, Rashmi B., Lee, Simon
• Primary Examiner(s): Spooner, Lamont M
• Application Number:
  US15/927,068
• Publication Number:
  US 20190317480A1
• Time in Patent Office: 2,177 Days
• CPC Class Codes: G05B 19/4184 characterised by fault tolerance, reliability of production system
• CPC Class Codes: G06F 16/35 Clustering; Classification
• CPC Class Codes: G06F 18/214 Generating training patterns; Bootstrap methods, e.g. bagging or boosting
• CPC Class Codes: G06F 18/24 Classification techniques
• CPC Class Codes: G06F 40/20 Natural language analysis semantic analysis of natural language G06F40/30
• CPC Class Codes: G06Q 10/06 Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
• CPC Class Codes: G06Q 10/20 Administration of product repair or maintenance
• CPC Class Codes: G06V 30/413 Classification of content, e.g. text, photographs or tables

Abstract

Some embodiments provide a program that retrieves a set of notifications describing failures that occurred on a set of monitored devices. The program further determines a set of topics based on the set of notifications. The program also determines failure modes associated with the set of topic from a plurality of failure modes defined for the set of monitored devices. The program further determines failure modes associated with the set of notifications based on the set of topics and the failure modes associated with the set of topics. The program also receives a particular notification that includes a particular set of words describing a failure that occurred on a particular monitored device in the set of monitored devices. The program further determines a failure mode associated with the particular notification based on the particular set of words and the determined failure modes associated with the set of notifications.

Claims

1. A non-transitory machine-readable medium storing a program executable by at least one processing unit of a device, the program comprising sets of instructions for:

generating labels for a labeled dataset, wherein generating the labels for the labeled dataset comprises;

retrieving a set of notifications from a storage, each notification comprising a set of words describing a failure that occurred on a monitored device in a set of monitored devices;

determining a set of topics based on the sets of words of the set of notifications, each topic being a distribution over words in the set of words;

for each topic in the set of topics, determining a failure mode from a plurality of failure modes defined for the set of monitored devices and associating the determined failure mode with the topic, wherein the failure mode associated with the topic is determined based on at least a similarity between the failure mode and the topic; and

for each notification in the set of notifications, determining a failure mode based on the set of topics and the failure modes associated with the set of topics and associating the determined failure mode with the notification, wherein the labels for the labeled dataset comprises the set of notifications and the determined failure modes associated with the set of notifications;

training a text classification model using the labeled dataset, wherein the text classification model is configured to determine failure modes from user-generated text describing failures;

receiving a particular notification from a client device, the particular notification comprising a particular set of words describing a failure that occurred on a particular monitored device in the set of monitored devices; and

determining, using the text classification model, a failure mode associated with the particular notification based on the particular set of words and the determined failure modes associated with the set of notifications.

2. The non-transitory machine-readable medium of claim 1, wherein a topic in the set of topics comprises a set of related words or phrases that occur in the set of words of the set of notifications.

3. The non-transitory machine-readable medium of claim 1, wherein determining the set of topics is further based on a Latent Dirichlet allocation (LDA) model.

4. The non-transitory machine-readable medium of claim 1, wherein determining, for each topic in the set of topics, the failure mode associated with the topic from the plurality of failure modes defined for the set of monitored devices is based on a Latent Dirichlet allocation (LDA) model.

5. The non-transitory machine-readable medium of claim 1, wherein determining, for each notification in the set of notifications, the failure mode associated with the notification comprises determining a topic in the set of topics associated with the notification and determining the failure mode associated with the notification based on the failure modes associated with the set of topics.

6. The non-transitory machine-readable medium of claim 5, wherein determining, for each notification in the set of notifications, the topic in the set of topics associated with the notification is based on a Latent Dirichlet allocation (LDA) model.

7. The non-transitory machine-readable medium of claim 1, wherein the storage is a first storage wherein determining, for each topic in the set of topics, the failure mode associated with the topic from the plurality of failure modes defined for the set of monitored devices comprises determining a device model associated with the set of monitored devices, determining that the plurality of failure modes are defined for the device model, and retrieving the plurality of failure modes from a second storage configured to store failure mode definitions.

8. A method comprising:

generating labels for a labeled dataset, wherein generating the labels for the labeled dataset comprises;

retrieving a set of notifications from a storage, each notification comprising a set of words describing a failure that occurred on a monitored device in a set of monitored devices;

determining a set of topics based on the sets of words of the set of notifications, each topic being a distribution over words in the set of words;

for each topic in the set of topics, determining a failure mode from a plurality of failure modes defined for the set of monitored devices and associating the determined failure mode with the topic, wherein the failure mode associated with the topic is determined based on at least a similarity between the failure mode and the topic; and

for each notification in the set of notifications, determining a failure mode based on the set of topics and the failure modes associated with the set of topics and associating the determined failure mode with the notification, wherein the labels for the labeled dataset comprises the set of notifications and the determined failure modes associated with the set of notifications;

training a text classification model using the labeled dataset, wherein the text classification model is configured to determine failure modes from user-generated text describing failures;

receiving a particular notification from a client device, the particular notification comprising a particular set of words describing a failure that occurred on a particular monitored device in the set of monitored devices; and

determining, using the text classification model, a failure mode associated with the particular notification based on the particular set of words and the determined failure modes associated with the set of notifications.

9. The method of claim 8, wherein a topic in the set of topics comprises a set of related words or phrases that occur in the set of words of the set of notifications.

10. The method of claim 8, wherein determining the set of topics is further based on a Latent Dirichlet allocation (LDA) model.

11. The method of claim 8, wherein determining, for each topic in the set of topics, the failure mode associated with the topic from the plurality of failure modes defined for the set of monitored devices is based on a Latent Dirichlet allocation (LDA) model.

12. The method of claim 8, wherein determining, for each notification in the set of notifications, the failure mode associated with the notification comprises determining a topic in the set of topics associated with the notification and determining the failure mode associated with the notification based on the failure modes associated with the set of topics.

13. The method of claim 12, wherein determining, for each notification in the set of notifications, the topic in the set of topics associated with the notification is based on a Latent Dirichlet allocation (LDA) model.

14. The method of claim 8, wherein the storage is a first storage wherein determining, for each topic in the set of topics, the failure mode associated with the topic from the plurality of failure modes defined for the set of monitored devices comprises determining a device model associated with the set of monitored devices, determining that the plurality of failure modes are defined for the device model, and retrieving the plurality of failure modes from a second storage configured to store failure mode definitions.

15. A system comprising:

a set of processing units; and

a non-transitory machine-readable medium storing instructions that when executed by at least one processing unit in the set of processing units cause the at least one processing unit to;

generate labels for a labeled dataset, wherein generating the labels for the labeled dataset comprises;

retrieving a set of notifications from a storage, each notification comprising a set of words describing a failure that occurred on a monitored device in a set of monitored devices;

determining a set of topics based on the sets of words of the set of notifications, each topic being a distribution over words in the set of words;

for each topic in the set of topics, determining a failure mode from a plurality of failure modes defined for the set of monitored devices and associating the determined failure mode with the topic, wherein the failure mode associated with the topic is determined based on at least a similarity between the failure mode and the topic; and

for each notification in the set of notifications, determining a failure mode based on the set of topics and the failure modes associated with the set of topics and associating the determined failure mode with the notification, wherein the labels for the labeled dataset comprises the set of notifications and the determined failure modes associated with the set of notifications;

train a text classification model using the labeled dataset, wherein the text classification model is configured to determine failure modes from user-generated text describing failures;

receive a particular notification from a client device, the particular notification comprising a particular set of words describing a failure that occurred on a particular monitored device in the set of monitored devices; and

determine, using the text classification model, a failure mode associated with the particular notification based on the particular set of words and the determined failure modes associated with the set of notifications.

16. The system of claim 15, wherein a topic in the set of topics comprises a set of related words or phrases that occur in the set of words of the set of notifications.

17. The system of claim 15, wherein determining the set of topics is further based on a Latent Dirichlet allocation (LDA) model.

18. The system of claim 15, wherein determining, for each topic in the set of topics, the failure mode associated with the topic from the plurality of failure modes defined for the set of monitored devices is based on a Latent Dirichlet allocation (LDA) model.

19. The system of claim 15, wherein determining, for each notification in the set of notifications, the failure mode associated with the notification comprises determining a topic in the set of topics associated with the notification and determining the failure mode associated with the notification based on the failure modes associated with the set of topics.

20. The system of claim 19, wherein determining, for each notification in the set of notifications, the topic in the set of topics associated with the notification is based on a Latent Dirichlet allocation (LDA) model.