Apparatus and a method for analyzing time series data for a plurality of items

• US 6,230,064 B1
• Filed: 06/29/1998
• Issued: 05/08/2001
• Est. Priority Date: 06/30/1997
• Status: Expired due to Fees

• Current Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Original Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Inventors: Kakimoto, Mitsuru, Nakase, Akihiko, Oyanagi, Shigeru
• Primary Examiner(s): Lee, Thomas
• Assistant Examiner(s): Mai, Rijue
• Application Number:
  US09/106,091
• Time in Patent Office: 1,044 Days
• Field of Search: 700/90-93, 345/339-380, 345/357, 705/8, 707/1-3
• US Class Current: 700/90
• CPC Class Codes: G06Q 10/06 Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling

Abstract

An association rule extraction apparatus extracts an association rule from time series data including events. A characteristic part extraction section extracts a characteristic change part from time series data of each event as an event sequence. The event sequence includes at least a start time of the characteristic change part as attribute data. An association rule extraction section extracts the association rule representing a tendency among the events in accordance with the attribute data.

Claims

1. An apparatus for analyzing time series data for a plurality of items, the time series data for each item consisting of values of variable arranged by a unit of a predetermined period, comprising:

characteristic part extraction means for extracting a characteristic change part of each item from each of a plurality of the time series data, each for the plurality of items, and for converting the characteristic change part to an event including an event name, a start time and a continuous time representing the characteristic change part, a first event of one item and a second event of another item for each time series data being an event pair; and

association rule extraction means for calculating for each event a distribution of the continuous time of each event for the same event name in each event pair, for calculating a distribution of the start time of one event for the same event name in each event pair if the each distribution of the continuous time satisfies a first standard, for calculating a distribution of a difference of the start time between the first event and the second event in each event pair if the distribution of the start time satisfies a second standard, and for determining the same event name and the difference of the start time as an association rule representing a tendency between the first event and the second event if the distribution of the difference of the start time satisfies a third standard.

2. The apparatus according to claim 1, wherein the event name is represented by a combination of a character change sign with a name of the item of the characteristic change part, and wherein the continuous time includes none.

3. The apparatus according to claim 1, wherein said characteristic part extraction means comprises:

means for extracting a maximum point and a minimum point from a graph corresponding to the time series data for each item,

means for creating a vector linked between the maximum point and the neighboring minimum point,

means for deciding whether an absolute value of inclination and a length of the vector are above each threshold, and

means for extracting a start point of the vector as the start time and the length of the vector as the continuous time if the absolute value and the length are above the each threshold.

4. The apparatus according to claim 1, wherein each of the plurality of the time series data each for the plurality of items is prepared by each client side.

5. The apparatus according to claim 1, wherein said association rule extraction means comprises:

means for deciding if the each distribution of the continuous time is below a first threshold as the first standard.

6. The apparatus according to claim 1, wherein said association rule extraction means comprises:

means for deciding if the distribution of the start time is above a second threshold as the second standard.

7. The apparatus according to claim 1,

wherein said association rule extraction means includes means for calculating a relative start time of the second event based on the start time of the first event in the each event pair.

8. The apparatus according to claim 1, wherein said association rule extraction means comprises:

means for deciding if the distribution of the relative start time in the each event pair is below a third threshold as the third standard.

9. The apparatus according to claim 6, wherein said association rule extraction means comprises:

means for deciding that the start time of the one event for the same event name relates to a predetermined time if the distribution of the start time is not above the second threshold.

10. A method for analyzing time series data for a plurality of items, the time series data for each item consisting of arranging values of variable by unit of a predetermined period, comprising the steps of:

extracting a characteristic change part of each item from each of a plurality of the time series data each for the plurality of items;

converting the characteristic change part to an event including an event name, a start time and a continuous time representing the characteristic change part, a first event of one item and a second event of another item for each time series data being an event pair;

calculating each distribution of the continuous time of each event for the same event name in each event pair;

calculating a distribution of the start time of one event for the same event name in each event pair if the each distribution of the continuous time satisfies a first standard;

calculating a distribution of a difference of the start time between the first event and the second event in each event pair if the distribution of the start time satisfies a second standard; and

determining the same event name and the difference of the start time as an association rule representing a tendency between the first event and the second event if the distribution of the difference of the start time satisfies a third standard.

11. The method according to claim 10, wherein the event name is represented by combining a character change sign with a name of the item of the characteristic change part, and wherein the continuous time includes none.

12. The method according to claim 10, wherein the step of extracting a characteristic change part comprises:

extracting a maximum point and a minimum point from a graph corresponding to the time series data for each item;

creating a vector linked between the maximum point and the neighboring minimum point;

deciding whether an absolute value of inclination and a length of the vector are above each threshold; and

extracting a start point of the vector as the start time and the length of the vector as the continuous time if the absolute value and the length are above each threshold.

13. The method according to claim 10, wherein each of the plurality of the time series data each for the plurality of items is prepared by each client side.

14. The method according to claim 10, wherein the step of calculating each distribution of the continuous time comprises:

deciding if the each distribution of the continuous time is below a first threshold as the first standard.

15. The method according to claim 10, wherein the step of calculating a distribution of the start time comprises:

deciding if the distribution of the start time is above a second threshold as the second standard.

16. The method according to claim 10, wherein the step of calculating a distribution of a difference comprises:

calculating a relative start time of the second event based on the start time of the first event in the each event pair.

17. The method according to claim 16, wherein the step of calculating a distribution of a difference comprises:

deciding if the distribution of the relative start time in the each event pair is below a third threshold as the third standard.

18. The method according to claim 15, wherein the step of calculating a distribution of the start time comprises:

deciding that the start time of the one event for the same event name relates to a predetermined time if the distribution of the start time is not above the second threshold.

19. A computer readable memory containing computer readable instructions to analyze time series data for a plurality of items, the time series data for each item consisting of arranging values of variable by unit of a predetermined period, comprising:

instruction means for causing a computer to extract a characteristic change part of each item from each of a plurality of the time series data each for the plurality of items;

instruction means for causing a computer to convert the characteristic change part to an event including an event name, a start time and a continuous time representing the characteristic change part, a first event of one item and a second event of another item for each time series data being an event pair;

instruction means for causing a computer to calculate each distribution of the continuous time of each event for the same event name in each event pair;

instruction means for causing a computer to calculate a distribution of the start time of one event for the same event name in each event pair if the each distribution of the continuous time satisfies a first standard;

instruction means for causing a computer to calculate a distribution of a difference of the start time between the first event and the second event in each event pair if the distribution of the start time satisfies a second standard; and

instruction means for causing a computer to determine the same event name and the difference of the start time as an association rule representing a tendency between the first event and the second event if the distribution of the difference of the start time satisfies a third standard.

20. A client apparatus for converting time series data to an event, comprising:

time series data memory for storing the time series data for the plurality of items, the time series data for each item consisting of arranging values of variable by unit of a predetermined period; and

characteristic part extraction means for extracting a characteristic change part of each item from the time series data, and for converting the characteristic change part to the event including an event name, a start time and a continuous time representing the characteristic change part, a first event of one item and a second event of another item for the time series data being an event pair.

21. A server apparatus for determining an association rule from a plurality of event pairs, comprising:

event memory for storing the plurality of event pairs, each event pair consisting of a first event of one item and a second event of another item for time series data of a plurality of items, the time series data of each item consisting of arranging values of variable by unit of a predetermined period, each event including an event name, a start time and a continuous time representing a characteristic change part, the characteristic change part of each item being extracted from the time series data of the plurality of items; and

association rule extraction means for calculating each distribution of the continuous time of each event for the same event name in each event pair, for calculating a distribution of the start time of one event for the same event name in each event pair if the each distribution of the continuous time satisfies a first standard, for calculating a distribution of a difference of the start time between the first event and the second event in each event pair if the distribution of the start time satisfies a second standard, and for determining the same event name and the difference of the start time as the association rule representing a tendency between the first event and the second event if the distribution of the difference of the start time satisfies a third standard.

22. A client method for converting time series data to an event, comprising the steps of:

storing the time series data of the plurality of items, the time series data of each item consisting of arranging values of variable by unit of a predetermined period;

extracting a characteristic change part of each item from the time series data; and

converting the characteristic change part to the event including an event name, a start time and a continuous time representing the characteristic change part, a first event of one item and a second event of another item for the time series data being an event pair.

23. A server method for determining an association rule from a plurality of event pairs, comprising the steps of:

storing the plurality of event pairs, each event pair consisting of a first event of one item and a second event of another item for time series data of a plurality of items, the time series data of each item consisting of arranging values of variable by unit of a predetermined period, each event including an event name, a start time and a continuous time representing a characteristic change part, the characteristic change part of each item being extracted from the time series data of the plurality of items;

calculating each distribution of the continuous time of each event for the same event name in each event pair;

calculating a distribution of the start time of one event for the same event name in each event pair if the each distribution of the continuous time satisfies a first standard;

calculating a distribution of a difference of the start time between the first event and the second event in each event pair if the distribution of the start time satisfies a second standard; and

determining the same event name and the difference of the start time as the association rule representing a tendency between the first event and the second event if the distribution of the difference of the start time satisfies a third standard.

24. A computer readable memory containing computer readable instructions to convert time series data to an event, comprising:

instruction means for causing a computer to store the time series data of the plurality of items, the time series data of each item consisting of arranging values of variable by unit of a predetermined period;

instruction means for causing a computer to extract a characteristic change part of each item from the time series data; and

instruction means for causing a computer to convert the characteristic change part to the event including an event name, a start time and a continuous time representing the characteristic change part, a first event of one item and a second event of another item for the time series data being an event pair.

25. A computer readable memory containing computer readable instructions to determine an association rule from a plurality of event pairs, comprising:

instruction means for causing a computer to store the plurality of event pairs, each event pair consisting of a first event of one item and a second event of another item for time series data of a plurality of items, the time series data of each item consisting of arranging values of variable by unit of a predetermined period, each event including an event name, a start time and a continuous time representing a characteristic change part, the characteristic change part of each item being extracted from the time series data of the plurality of items;

instruction means for causing a computer to calculate each distribution of the continuous time of each event for the same event name in each event pair;

instruction means for causing a computer to calculate a distribution of the start time of one event for the same event name in each event pair if the each distribution of the continuous time satisfies a first standard;

instruction means for causing a computer to calculate a distribution of a difference of the start time between the first event and the second event in each event pair if the distribution of the start time satisfies a second standard; and

instruction means for causing a computer to determine the same event name and the difference of the start time as the association rule representing a tendency between the first event and the second event if the distribution of the difference of the start time satisfies a third standard.