CAUSAL MODELING OF MULTI-DIMENSIONAL HIERACHICAL METRIC CUBES

• US 20120116850A1
• Filed: 11/10/2010
• Published: 05/10/2012
• Est. Priority Date: 11/10/2010
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Abe, Naoki, Kapoor, Shubir, Lozano, Aurelie C., Ray, Bonnie K., Fu, Jing, Gemmell, Michael G., Kelly, Floyd S., Loehr, David M., Mahatma, Shilpa N.
• Granted Patent: US 10,360,527 B2
• Time in Patent Office: Days
• US Class Current: 705/7.38
• CPC Class Codes: G06Q 10/063 Operations research, analysis or management
• CPC Class Codes: G06Q 10/0639 Performance analysis of employees; Performance analysis of enterprise or organisation operations
• CPC Class Codes: G06Q 10/067 Enterprise or organisation modelling

Abstract

A computing system initializes a first frontier to be a root of a multi-dimensional hierarchical data structure representing an entity. The system acquires first data corresponding to the first frontier. The system performs modeling on the first data to obtain a first model and a corresponding first statistic. The system expands a dimension of the first frontier. The system gathers second data corresponding to the expanded frontier. The system applies the data modeling on the second data to obtain a second model and a corresponding second statistic. The system compares the first statistic of the first model and the second statistic of the second model. The system sets the second model to be the first model in response to determining that the second model statistic is better than the first model statistic. The system outputs the first model.

Claims

1. A method for managing performance of an entity, the method comprising:

initializing a first frontier to a root of a multi-dimensional hierarchical data structure representing the entity, the frontier referring to a cut of the multi-dimensional hierarchical data structure at which modeling of at least one metric of the entity is performed;

acquiring first data corresponding to the first frontier;

performing modeling on the first data to obtain a first model and a corresponding first statistic;

expanding a dimension of the frontier to obtain a new frontier;

gathering second data corresponding to the expanded frontier;

applying the data modeling on the second data to obtain a second model and a corresponding second statistic;

comparing the first statistic of the first model and the second statistic of the second model;

setting the second model to be the first model in response to determining that the second statistic is better than the first model statistic; and

outputting the first model,

wherein a processor performs one or more of;

the initializing, the acquiring, the performing, the expanding, the gathering, the applying, the comparing, the setting and the outputting.

2. The method according to claim 1, wherein the modeling includes one or more of:

probabilistic graphical modeling, least squares estimation, graphical Granger modeling method, and generalized linear modeling method.

3. The method according to claim 1, wherein the acquiring the first data includes one or more of:

aggregating data corresponding to nodes in a dimension of the hierarchical data structure;

or

constructing a separate model corresponding to distinct nodes in one or more dimensions of the hierarchical data structure.

4. The method according to claim 1, wherein the modeling includes:

specifying constraints in relationships between metrics, the constraints restricting the first model and the second model.

5. The method according to claim 1, wherein the comparing includes:

calculating, as the first and second statistics that represent statistical values, respective statistical significances of the first model and the second model, and comparing the calculated statistical significances,

wherein the second model is set to be the first model in response to determining that the second model statistic is higher than the first model statistic.

6. The method according to claim 1, wherein the comparing includes:

calculating, as the first and second statistics that represent statistical values, respective predictive squared error means of the first model and the second model, and comparing the calculated predictive squared error means.

7. The method according to claim 1, wherein the outputted first model is used in a first analysis, the first analysis comprises:

optionally setting values of a set of metrics in the outputted first model to new values;

optionally allocating or aggregating the first data to obtain the second data;

performing prediction on the values of metrics by using the outputted first model and optionally set values; and

allocating or aggregating predicted values at a level lower or higher than the new frontier.

8. The method according to claim 7, further comprising:

analyzing the output model based on a second analysis, the second analysis determining a root cause of a deviation or anomaly in the values of the metrics,

wherein the second analysis comprises;

performing a prediction according to which some of the metrics in the data structure are changed towards mean values of the some of the metrics;

examining an effect of the changes in the values of the metrics; and

determining which metric at a particular time is to be attributed as a main cause of the deviation in the metrics by identifying which metrics result in bringing back the value of the metrics to normal value range with minimal changes to the values.

9. The method according to claim 7, further comprising:

analyzing the output model based on a third analysis, the third analysis determining a set of candidate metrics for which a change in a value of the candidate metrics causes a predicted value of a target metric to achieve a desired value;

wherein the third analysis comprises;

performing a prediction, in which metrics in the data structure are changed in a direction suggested by the output model;

examining the effect of said changes to the values of the candidate metrics; and

determining which metric in the data structure is projected to achieve the desired value of the target metric, with minimal changes to the values of the candidate metrics.

10. The method according to claim 1, further comprising:

determining a measure of strength of causal relations identified in the outputted first model by calculating for each of the causal relations a score that is a function of model coefficients associated with the said causal relation, the score including one or more of;

L1 norm, L2 norm, sum, and maximum.

11. The method according to claim 10, further comprising:

determining causal relations and associated measures of strengths at cuts in the frontier in the outputted first model by calculating a causal strength, per each candidate relation or a pair of metrics, in the cuts in the frontier, as a weighted sum of causal strengths of all causal relations included in the outputted first model, whose dimension nodes are equal or descendents of a candidate relation, in which a weight of each causal strength is determined by a ratio between a value of a target metric at the frontier and an aggregated value of the target metric at the cuts.

12. The method according to claim 1, further comprising:

repeating the expanding, gathering, applying, comparing, setting and outputting to expand at least one dimension of the first frontier that has not yet been expanded, if the second statistic is not better than the first statistic and if there exists the at least one dimension.

13. A system for managing performance of an entity, the system comprising:

a memory device; and

a processor being connected to the memory device,

wherein the processor performs steps of;

initializing a first frontier to a root of a multi-dimensional hierarchical data structure representing the entity, the frontier referring to a cut of the multi-dimensional hierarchical data structure at which modeling of at least one metric of the entity is performed;

acquiring first data corresponding to the first frontier;

performing modeling on the first data to obtain a first model and a corresponding first statistic;

expanding a dimension of the frontier to obtain a new frontier;

gathering second data corresponding to the expanded frontier;

applying the data modeling on the second data to obtain a second model and a corresponding second statistic;

comparing the first statistic of the first model and the second statistic of the second model;

setting the second model to be the first model in response to determining that the second model statistic is better than the first model statistic; and

outputting the first model.

14. The system according to claim 13, wherein the modeling includes one or more of:

probabilistic graphical modeling, least squares estimation, graphical Granger modeling method, and generalized linear modeling.

15. The system according to claim 13, wherein the acquiring the first data includes one or more of:

aggregating data corresponding to nodes in a dimension of the hierarchical data structure;

or

constructing a separate model corresponding to distinct nodes in one or more of the hierarchical data structure.

16. The system according to claim 13, wherein the modeling includes:

specifying constraints on relationships between metrics, the constraints restricting the first model and the second model.

17. The system according to claim 13, wherein the comparing includes:

calculating, as the first and second statistics that represent statistical values, respective statistical significances of the first model and the second model, and comparing the calculated statistical significances,

wherein the second model is set to be the first model in response to determining that the second model statistic is higher than the first model statistic.

18. The system according to claim 13, wherein the comparing includes:

calculating, as the first and second statistics that represent statistical values, respective predictive squared error means of the first model and the second model, and comparing the calculated predictive squared error means.

19. The system according to claim 13, wherein the outputted first model is used in a first analysis, the first analysis comprises:

optionally setting values of a set of metrics in the outputted first model to new values;

optionally allocating or aggregating the first data to obtain the second data;

performing prediction on the values of metrics by using the outputted first model and the optionally set values; and

allocating or aggregating predicted values at a level lower or higher than the new frontier.

20. The system according to claim 19, the processor further performs a step of:

analyzing the outputted first model based on a second analysis, the second analysis determining a root cause of a deviation or anomaly in the values of the metrics,

wherein the second analysis comprises;

performing a prediction according to which some of the metrics in the data structure are changed towards mean values of the some of the metrics;

examining an effect of the changes in the values of the metrics; and

determining which metric at a particular time is to be attributed as a main cause of the deviation in the metrics by identifying which metrics result in bringing back the value of the metrics to normal value range with minimal changes to the values.

21. The system according to claim 20, the processor further performs a step of:

analyzing the output first model based on a third analysis, the third analysis determining a set of candidate metrics for which a change in a value of the candidate metrics causes a predicted value of a target metric to achieve a desired value;

wherein the third analysis comprises;

performing a prediction, in which metrics in the data structure are changed in a direction suggested by the output model;

examining the effect of said changes to the values of the candidate metrics; and

determining which metric in the data structure is projected to achieve the desired value of the target metric, with minimal changes to the values of the candidate metrics.

22. The system according to claim 13, wherein the processor further performs a step of:

determining a measure of strength of causal relations identified in the outputted first model by calculating for each of the causal relations a score that is a function of model coefficients associated with the said causal relation, the score including one or more of;

L1 norm, L2 norm, sum, and maximum.

23. A computer program product for managing performance of an entity, the computer program product comprising a storage medium readable by a processing circuit and storing instructions run by the processing circuit for performing a method, the method comprising:

initializing a first frontier to a root of a multi-dimensional hierarchical data structure representing the entity, the frontier referring to a cut of the multi-dimensional hierarchical data structure at which modeling of at least one metric of the entity is performed;

acquiring first data corresponding to the first frontier;

performing data modeling on the first data to obtain a first model and a corresponding first statistic;

expanding a dimension of the frontier to obtain a new frontier;

gathering second data corresponding to the expanded frontier;

applying the modeling on the second data to obtain a second model and a corresponding second statistic;

comparing the first statistic of the first model and the second statistic of the second model;

setting the second model to be the first model in response to determining that the second model statistic is better than the first model statistic; and

outputting the first model.

24. The computer program product according to claim 23, wherein the comparing includes:

calculating, as the first and second statistics that represent statistical values, respective statistical significances of the first model and the second model, and comparing the calculated statistical significances, wherein the second model is set to be the first model in response to determining that the second model statistic is higher than the first model statistic.