Computer-implemented systems and methods for forecast reconciliation

• US 8,112,302 B1
• Filed: 08/31/2007
• Issued: 02/07/2012
• Est. Priority Date: 11/03/2006
• Status: Active Grant

• Current Assignee: SAS Institute Incorporated
• Original Assignee: SAS Institute Incorporated
• Inventors: Trovero, Michele Angelo, Joshi, Mahesh V., Leonard, Michael James, Fahey, Richard Patrick, Golovashkin, Dmitry V.
• Primary Examiner(s): Nguyen, Nga B
• Assistant Examiner(s): MARCUS, LELAND R
• Application Number:
  US11/848,982
• Time in Patent Office: 1,621 Days
• Field of Search: 705/7
• US Class Current: 705/7.31
• CPC Class Codes: G06Q 10/00 Administration; Management
• CPC Class Codes: G06Q 10/063 Operations research, analysis or management
• CPC Class Codes: G06Q 30/0201 Market modelling; Market analysis; Collecting market data
• CPC Class Codes: G06Q 30/0202 Market predictions or forecasting for commercial activities

Abstract

Systems and methods for reconciling a forecast for a dimension based upon data that is associated with the dimension. A method can be used that includes generating a plurality of forecasts for the dimensions such that the forecast of a first dimension is generated independently of a forecast of a second dimension. The forecast of the first dimension has a constraint that is influenced by the forecast of the second dimension. A reconciliation is performed between the forecast of the first dimension and the forecast of the second dimension in order to determine how the constraint of the first dimension'"'"'s forecast is to influence the first dimension'"'"'s forecast.

Claims

1. A computer-implemented method for reconciling forecasts, comprising:

determining constraints, using one or more data processors, wherein the constraints include one or more explicit constraints and one or more implicit constraints for forecasts to be generated, wherein an implicit constraint includes a requirement that a reconciled forecast for a parent node is based upon an aggregation of one or more reconciled forecasts for children nodes of the parent node, wherein the reconciled forecast for the parent node is mathematically equivalent to the aggregation of the reconciled forecasts for the children nodes, and wherein the reconciled forecasts include associated measures of uncertainty;

generating, using the one or more data processors, forecasts for a plurality of nodes hierarchically arranged in parent-child relationships, wherein the generated forecasts are independently generated for each node without consideration of other forecasts; and

using, using the one or more data processors, the one or more explicit and implicit constraints and a non-linear optimizer to reconcile the generated forecasts, wherein the non-linear optimizer adjusts one or more of the generated forecasts so that the constraints are met while concurrently minimizing a loss function that penalizes a functional of the differences between reconciled forecasts and generated forecasts.

2. The method of claim 1, wherein one of the explicit constraints is a judgmental forecast or bound imposed by a software system.

3. The method of claim 2, wherein the one of the explicit constraints is a linear or non-linear based constraint.

4. The method of claim 1 further comprising:

determining, using the one or more processors, whether a constraint conflict exists among the constraints; and

receiving, using the one or more processors, conflict resolution information to resolve an identified constraint conflict.

5. The method of claim 4, wherein one of the explicit constraints is a judgmental forecast, wherein a conflict arises because the judgmental forecast conflicts with another constraint, wherein the conflict results in an infeasible reconciliation problem, and wherein the conflict resolution information resolves the conflict.

6. The method of claim 1, wherein the generated forecasts are generated based upon a plurality of forecast series related to product information, and wherein the forecast series are organized into a hierarchy by a product characteristic and by a geographical location.

7. The method of claim 1, wherein one of the explicit constraints is a bound constraint that specifies acceptable values or ranges corresponding to a generated forecast.

8. The method of claim 1, wherein one of the explicit constraints is a judgmental constraint or override.

9. The method of claim 1, wherein the independently generated forecasts do not take the one or more implicit constraints into account when being generated, and wherein the non-linear optimizer enforces the one or more implicit constraints such that the reconciled forecasts do align with one or more implicit constraints.

10. The method of claim 1, wherein the loss function penalizes a difference between a reconciled forecast and a generated forecast for a first node more than a difference between a reconciled forecast and a generated forecast for a second node.

13. The method of claim 1, wherein the non-linear optimizer generates the reconciled forecasts top-down, bottom-up, or middle-out.

14. The method of claim 1, wherein the nodes are arranged in levels, wherein a first level is associated with particular products, wherein a second level is associated with product lines, and wherein a third level is associated with regions.

15. The method of claim 10, wherein the forecast for the first node is more reliable than the forecast for the second node.

16. The method of claim 1, wherein explicit constraints are associated with fewer than all of the generated forecasts.

17. The method of claim 1, wherein the implicit constraint includes a requirement that a reconciled forecast for a parent node is a sum or average of the one or more reconciled forecasts for children nodes of the parent node.

18. The method of claim 1, wherein the constraints are met while concurrently minimizing the loss function.

19. The method of claim 1, wherein minimizing the loss function includes minimizing a quadratic loss function using an efficient interior point programming algorithm.

20. The method of claim 1, wherein the generated forecasts are time series forecasts and the reconciled forecasts are time series forecasts.

21. The method of claim 1, wherein the generated forecasts are generated using a stochastic process, and the generated forecasts are adjusted using a deterministic process to generate the reconciled forecasts.

24. The method of claim 1, wherein reconciled forecasts are provided to a decision process system for analysis.

11. A computer-implemented system for reconciling forecasts, comprising:

one or more processors;

one or more computer-readable storage media containing instructions configured to cause the one or more processors to perform operations including;

determining constraints, wherein the constraints include one or more explicit constraints and one or more implicit constraints for forecasts to be generated, wherein an implicit constraint includes a requirement that a reconciled forecast for a parent node is based upon an aggregation of one or more reconciled forecasts for children nodes of the parent node, wherein the reconciled forecast for the parent node is mathematically equivalent to the aggregation of the reconciled forecasts for the children nodes, and wherein the reconciled forecasts include associated measures of uncertainty;

generating forecasts for a plurality of nodes hierarchically arranged in parent-child relationships, wherein the generated forecasts are independently generated for each node without consideration of other forecasts; and

using the constraints and a non-linear optimizer to reconcile the generated forecasts, wherein the non-linear optimizer adjusts one or more of the generated forecasts so that the constraints are met while concurrently minimizing a loss function that penalizes differences between reconciled forecasts and generated forecasts.

12. A computer program product for providing row-level security, tangibly embodied in a machine-readable storage medium, including instructions configured to cause a data processing system to:

determine constraints, wherein the constraints include one or more explicit constraints and one or more implicit constraints for forecasts to be generated, wherein an implicit constraint includes a requirement that a reconciled forecast for a parent node is based upon an aggregation of one or more reconciled forecasts for children nodes of the parent node, wherein the reconciled forecast for the parent node is mathematically equivalent to the aggregation of the reconciled forecasts for the children nodes, and wherein the reconciled forecasts include associated measures of uncertainty;

generate forecasts for a plurality of nodes hierarchically arranged in parent-child relationships, wherein the generated forecasts are independently generated for each node without consideration of other forecasts; and

use the constraints and a non-linear optimizer to reconcile the generated forecasts, wherein the non-linear optimizer adjusts one or more of the generated forecasts so that the constraints are met while concurrently minimizing a loss function that penalizes differences between reconciled forecasts and generated forecasts.

22. A computer-implemented method for reconciling forecasts, comprising:

generating, using one or more data processors, forecasts for a plurality of nodes hierarchically arranged in parent-child relationships, wherein the generated forecasts are independently generated for each node without consideration of other forecasts;

determining constraints, using the one or more data processors, wherein the constraints include one or more explicit constraints and one or more implicit constraints, wherein an implicit constraint includes a requirement that a reconciled forecast for a parent node is based upon an aggregation of one or more reconciled forecasts for children nodes of the parent node, wherein the reconciled forecast for the parent node is mathematically equivalent to the aggregation of the reconciled forecasts for the children nodes, and wherein the reconciled forecasts include associated measures of uncertainty; and

using, using the one or more data processors, the constraints and a non-linear optimizer to reconcile the generated forecasts, wherein the non-linear optimizer adjusts one or more of the generated forecasts so that the constraints are met while concurrently minimizing a loss function that penalizes differences between reconciled forecasts and generated forecasts.

23. A computer-implemented method for reconciling forecasts, comprising:

generating, using one or more data processors, forecasts for a plurality of nodes hierarchically arranged in parent-child relationships, wherein the generated forecasts are independently generated for each node without consideration of other forecasts;

wherein constraints, including one or more explicit constraints and one or more implicit constraints, are determined for the generated forecasts, wherein the constraints are used to reconcile the generated forecasts, wherein an implicit constraint includes a requirement that a reconciled forecast for a parent node is based upon an aggregation of one or more reconciled forecasts for children nodes of the parent node, wherein the reconciled forecast for the parent node is mathematically equivalent to the aggregation of the reconciled forecasts for the children nodes, and wherein the reconciled forecasts include associated measures of uncertainty; and

using, using the one or more data processors, a non-linear optimizer to generate reconciled forecasts, wherein the non-linear optimizer adjusts one or more of the generated forecasts so that the constraints are met while minimizing a loss function that penalizes differences between reconciled forecasts and generated forecasts.