PROACTIVE EVENT DRIVEN COMPUTING

• US 20140019398A1
• Filed: 07/15/2012
• Published: 01/16/2014
• Est. Priority Date: 07/15/2012
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Feldman, Zohar, Engel, Yagil
• Granted Patent: US 10,643,133 B2
• Time in Patent Office: Days
• US Class Current: 706/46
• CPC Class Codes: G06N 5/02 Knowledge representation; Symbolic representation
• CPC Class Codes: G06Q 10/0635 Risk analysis of enterprise or organisation activities

Abstract

A method for predicting a future situation based on an analysis of at least one predictive pattern. The method comprises monitoring a plurality of events carried out by an event processing component, detecting a predictive pattern predictive of a future situation, selecting one of a plurality of proactive actions and an execution time according to its effect on at least one of a probability of occurrence and a cost of occurrence of the future situation, and outputting the selected proactive action.

Claims

1. A method for predicting a future situation based on an analysis of at least one predictive pattern, comprising:

monitoring a plurality of events carried out by an event processing component;

detecting a predictive pattern predictive of a future situation;

selecting one of a plurality of proactive actions and an execution time for said selected proactive action according to its effect when executed in said execution time on at least one of a probability of occurrence and a cost of occurrence of said future situation; and

outputting said selected proactive action.

2. The method of claim 1, wherein said selecting is performed according to a rule defining said selected proactive action as a response to said predictive pattern.

3. The method of claim 1, wherein said selecting comprises calculating for each said proactive action in at least one execution time a combination between said probability of occurrence of said future situation and said cost of occurrence of said future situation and performing said selecting according to said calculating.

4. The method of claim 1, wherein said cost of occurrence changes as a function of time during a period lasting until said future situation is estimated to occur, said selecting comprises calculating said effect in a plurality of instances during said period.

5. The method of claim 1, wherein said probability of occurrence changes as a function of time during a period lasting until said future situation is estimated to occur, said selecting comprises calculating said effect in a plurality of instances during said period.

6. The method of claim 1, wherein said cost changes as a function of time during a decision epoch.

7. The method of claim 1, wherein said probability of occurrence dynamically changes during said monitoring while said future situation is continuously calculated as a function of at least one dynamic characteristic of said event processing component and dynamic parameter that effects said event processing component.

8. The method of claim 1, wherein said predictive pattern defines at least one event marked with a measure of probability of event occurrence said probability of occurrence is calculated according to said measure of said at least one event.

9. The method of claim 1, wherein said selecting comprises calculating said effect for each said proactive action in real time, in response to said detecting.

10. The method of claim 1, wherein said selecting comprises estimating said effect for each said proactive action in a plurality of future instances;

wherein said selecting comprises selecting in which of said plurality of future instances to perform said selected proactive action.

11. The method of claim 10, wherein said cost of occurrence calculated by taking into account a cost of performing said selected proactive action at a respective said future instance.

12. The method of claim 1, wherein said cost of occurrence is a combination of a first cost of performing said selected proactive action and a second cost of the effect of said future situation on the activity of said event processing component after said selected proactive action is performed.

13. The method of claim 1, wherein said probability of occurrence is a probability distribution over an occurrence time of said future situation conditioned on the detection of said predictive pattern.

14. The method of claim 1, wherein said selecting comprises applying a Markov decision process (MDP) on said plurality of proactive actions.

15. The method of claim 14, wherein said MDP is solved using a backwards induction.

16. The method of claim 1, wherein said selected proactive action is a combination of a plurality of proactive actions.

17. A computer readable medium comprising computer executable instructions adapted to perform the method of claim 1.

18. A system of predicting a future situation based on an analysis of at least one predictive pattern, comprising:

a processor;

a monitoring which monitors an event processing component carrying out a plurality of events;

a detection module which detects a predictive pattern predictive of a future situation; and

prediction module which identifies a plurality of proactive actions each having an effect on at least one of a probability of occurrence and a cost of occurrence of said future situation and selects, using said processor, one of said plurality of proactive actions according to said effect.

19. A computer program product for predicting a future situation based on an analysis of at least one predictive pattern, comprising:

a computer readable storage medium;

first program instructions to monitor a plurality of events carried out by an event processing component;

second program instructions to detect a predictive pattern predictive of a future situation; and

third program instructions to select one of a plurality of proactive actions and an execution time according to its effect on at least one of a probability of occurrence and a cost of occurrence of said future situation;

wherein said first, second, and third program instructions are stored on said computer readable storage medium.

20. A method for predicting a future situation based on an analysis of at least one predictive pattern, comprising:

monitoring a plurality of events carried out by an event processing component;

detecting a predictive pattern predictive of a future situation;

identifying a predefined rule that defines which of a plurality of proactive actions to take in response to said detection from a plurality of predefined rules; and

performing said proactive action in response to said detection;

wherein said predefined rule is automatically calculated according to an effect of said proactive action on at least one of a probability of occurrence and a cost of occurrence of said future situation in relation to the effect of other said plurality of proactive actions.