Rules generation for IT resource event situation classification

• US 7,895,137 B2
• Filed: 07/17/2009
• Issued: 02/22/2011
• Est. Priority Date: 04/27/2005
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Minerley, Kevin Gordon, Sheng, Ma, Loewenstern, David Matthew, Salahshour, Abdolreza
• Primary Examiner(s): Sparks; Donald
• Assistant Examiner(s): Brown, Jr.; Nathan H
• Application Number:
  US12/505,212
• Publication Number:
  US 20090276383A1
• Time in Patent Office: 585 Days
• Field of Search: 706/12
• US Class Current: 706/12
• CPC Class Codes: G06F 9/542 Event management; Broadcasting; Multicasting; Notifications

Abstract

A computer processing device receives computer readable data to derive computer executable rules for mining and constructing situation categories. The received data is transformed into a predetermined standard format if the received data is not already in the predetermined standard format. The predetermined standard formatted data is parsed, and an outer, iterative loop is performed until at least one predetermined stopping criterion is met. An inner iterative loop is performed within the outer iterative loop until all desired subsets of data are processed. During the inner iterative loop, selected subsets of data are labeled with labels associated with corresponding previously labeled subsets of data. New computer executable rules are generated for mining and constructing situation categories from the labeled subsets of data. Keyword list classifiers are transformed using the stored labeled subsets of data.

Claims

1. A method of transforming data into computer executable rules for mining and constructing situation categories that are applied to information technology resource messages or events comprising:

receiving computer readable data by a computer processing device from at least one of;

a raw log and a catalog, where the received data is at least one of;

initial seed data and knowledge data, to derive the computer executable rules for mining and constructing situation categories;

transforming the received data into a predetermined standard format if the received data is not already in the predetermined standard format;

parsing the predetermined standard formatted data;

performing an outer, iterative loop until at least one predetermined stopping criterion is met, comprising;

utilizing a keyword rule classifier by the computer processing device to automatically pre-classify at least a portion of the parsed data;

performing an inner iterative loop within the outer iterative loop, comprising;

selecting a subset of the parsed data for expert review;

using at least one of keyword rules, features, and classifications to find, within data available to the computer processing device, a corresponding previously labeled subset of data that has similar semantics to semantics of the selected subset of data;

labeling the selected subset of data with the label associated with the corresponding previously labeled subset of data; and

repeating the inner iterative loop if another subset of data is to be processed;

storing each labeled subset of data on a data storage device;

generating new computer executable rules for mining and constructing situation categories from the stored labeled subsets of data;

transforming keyword list classifiers using the stored labeled subsets of data; and

repeating the outer iterative loop if the predetermined stopping criterion is not met.

2. The method of claim 1, wherein labeling the selected subset of data comprises re-labeling previously labeled data in the selected subset of data.

3. The method of claim 1, further comprising:

creating an event to situation catalog if the source of the received computer readable data is from a catalog.

4. The method of claim 1, wherein:

parsing the predetermined standard formatted data comprises parsing the predetermined standard formatted data into event entries, each event entry comprising at least one field of event data;

further comprising;

preprocessing the parsed data within the outer iterative loop by performing featurization of the parsed data to break the parsed data down into features that can be used to build and match rules by;

selecting fields from the parsed data;

identifying fields that are considered natural language fields;

identifying tokenized fields from the fields also identified as natural language fields;

transforming identified tokenized fields into features by attaching a field identifier to each tokenized field;

transforming fields that are not considered natural language fields into a single feature by attaching a field identifier to the content of the transformed field;

tagging features with a property name so that features from different properties are handled separately; and

storing in the computer readable storage, two representations for each event entry, including a first representation that pairs each feature with its original token and a second representation as an unordered set of features.

5. The method of claim 4, further comprising:

performing at least one of;

combining fields before featurization;

keeping features from different selected fields separate; and

editing features to reduce the number of features.

6. The method of claim 1, wherein using at least one of keyword rules, features, and classifications to find, within data available to the computer processing device, a corresponding previously labeled subset of data comprises a user-expert selecting a subset of the data and using a combination of keyword rules, features, and classifications to find a previously labeled subset of data that the user-expert determines has similar semantics to semantics of the selected subset of data.

7. The method of claim 1, wherein using at least one of keyword rules, features, and classifications to find, within data available to the computer processing device, a corresponding previously labeled subset of data comprises determining if all selected features of previously labeled subsets of data correspond to a feature set of the selected subset of data and determining if no negated features of the previously labeled subset of data appear in the feature set of the selected subset of data.

8. The method of claim 1, wherein generating new computer executable rules for mining and constructing situation categories from the stored labeled subsets of data comprises generating rules including a pattern and a corresponding situation category.

9. The method of claim 8, wherein each pattern is composed of two sets of features, including inclusion features and exclusion features, such that a selected rule matches an event if all of the inclusion features of the selected rule appear in the feature set of the entry and none of the exclusion features of the selected rule appear in the feature set of the entry.

10. The method of claim 1, wherein generating new computer executable rules comprises implementing a computer executable classification algorithm to generate candidate rules and a heuristic filter to remove selected ones of the candidate rules judged to be bad candidate rules.

11. A computer program product for transforming data into computer executable rules for mining and constructing situation categories that are applied to information technology resource messages or events comprising:

a computer usable medium having computer usable program code embodied therewith, the computer usable program code comprising;

computer usable program code configured to receive computer readable data by a computer processing device from at least one of;

a raw log and a catalog, where the received data is at least one of;

initial seed data and knowledge data, to derive the computer executable rules for mining and constructing situation categories;

computer usable program code configured to transform the received data into a predetermined standard format if the received data is not already in the predetermined standard format;

computer usable program code configured to parse the predetermined standard formatted data;

computer usable program code configured to perform an outer, iterative loop until at least one predetermined stopping criterion is met, comprising;

computer usable program code configured to utilize a keyword rule classifier by the computer processing device to automatically pre-classify at least a portion of the parsed data;

computer usable program code configured to perform an inner iterative loop within the outer iterative loop, comprising;

computer usable program code configured to select a subset of the parsed data for expert review;

computer usable program code configured to use at least one of keyword rules, features, and classifications to find, within data available to the computer processing device, a corresponding previously labeled subset of data that has similar semantics to semantics of the selected subset of data;

computer usable program code configured to label the selected subset of data with the label associated with the corresponding previously labeled subset of data; and

computer usable program code configured to repeat the inner iterative loop if another subset of data is to be processed;

computer usable program code configured to store each labeled subset of data on a data storage device;

computer usable program code configured to generate new computer executable rules for mining and constructing situation categories from the stored labeled subsets of data;

computer usable program code configured to transform keyword list classifiers using the stored labeled subsets of data; and

computer usable program code configured to repeat the outer iterative loop if the predetermined stopping criterion is not met.

12. The computer program product of claim 11, wherein computer usable program code configured to label the selected subset of data comprises computer usable program code configured to re-label previously labeled data in the selected subset of data.

13. The computer program product of claim 11, further comprising:

computer usable program code configured to create an event to situation catalog if the source of the received computer readable data is from a catalog.

14. The computer program product of claim 11, wherein:

computer usable program code configured to parse the predetermined standard formatted data comprises computer usable program code configured to parse the predetermined standard formatted data into event entries, each event entry comprising at least one field of event data;

further comprising;

computer usable program code configured to preprocess the parsed data within the outer iterative loop by performing featurization of the parsed data to break the parsed data down into features that can be used to build and match rules comprising;

computer usable program code configured to select fields from the parsed data;

computer usable program code configured to identify fields that are considered natural language fields;

computer usable program code configured to identify tokenized fields from the fields also identified as natural language fields;

computer usable program code configured to transform identified tokenized fields into features by attaching a field identifier to each tokenized field;

computer usable program code configured to transform fields that are not considered natural language fields into a single feature by attaching a field identifier to the content of the transformed field;

computer usable program code configured to tag features with a property name so that features from different properties are handled separately; and

computer usable program code configured to store in the computer readable storage, two representations for each event entry, including a first representation that pairs each feature with its original token and a second representation as an unordered set of features.

15. The computer program product of claim 14, further comprising:

computer usable program code configured to perform at least one of;

combining fields before featurization;

keeping features from different selected fields separate; and

editing features to reduce the number of features.

16. The computer program product of claim 11, wherein computer usable program code configured to use at least one of keyword rules, features, and classifications to find, within data available to the computer processing device, a corresponding previously labeled subset of data comprises computer usable program code configured to implement a user interface to allow a user-expert to select a subset of the data and use a combination of keyword rules, features, and classifications to find a previously labeled subset of data that the user-expert determines has similar semantics to semantics of the selected subset of data.

17. The computer program product of claim 11, wherein computer usable program code configured to use at least one of keyword rules, features, and classifications to find, within data available to the computer processing device, a corresponding previously labeled subset of data comprises computer usable program code configured to determine if all selected features of previously labeled subsets of data correspond to a feature set of the selected subset of data and determine if no negated features of the previously labeled subset of data appear in the feature set of the selected subset of data.

18. The computer program product of claim 11, wherein computer usable program code configured to generate new computer executable rules for mining and constructing situation categories from the stored labeled subsets of data comprises computer usable program code configured to generate rules including a pattern and a corresponding situation category.

19. The computer program product according to claim 18, wherein each pattern is composed of two sets of features, including inclusion features and exclusion features, such that a selected rule matches an event if all of the inclusion features of the selected rule appear in the feature set of the entry and none of the exclusion features of the selected rule appear in the feature set of the entry.

20. The computer program product according to claim 11, wherein computer usable program code configured to generate new computer executable rules comprises computer usable program code configured to implement a computer executable classification algorithm to generate candidate rules and a heuristic filter to remove selected ones of the candidate rules judged to be bad candidate rules.