CONTEXT-AWARE PARAMETER ESTIMATION FOR FORECAST MODELS

• US 20130024170A1
• Filed: 07/20/2012
• Published: 01/24/2013
• Est. Priority Date: 07/21/2011
• Status: Active Grant

• Current Assignee: SAP SE
• Original Assignee: SAP AG (SAP SE)
• Inventors: Lehner, Wolfgang, Dannecker, Lars, Schulze, Robert, Boehm, Matthias
• Granted Patent: US 9,361,273 B2
• Time in Patent Office: Days
• US Class Current: 703/2
• CPC Class Codes: G06F 17/18 for evaluating statistical data , e.g. average values, frequency distributions, probability functions, regression analysis forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04
• CPC Class Codes: G06F 2119/06 Power analysis or power optimisation
• CPC Class Codes: G06F 30/20 Design optimisation, verification or simulation optimisation, verification or simulation of circuit designs G06F30/30
• CPC Class Codes: G06Q 10/063 Operations research, analysis or management

Abstract

Methods, systems, and computer-readable storage media for providing at least one parameter for use with a forecast model. Implementations include actions of receiving a first context vector, the first context vector including a plurality of context attributes that describe a first context, retrieving a first parameter vector from a repository based on the first context vector, the repository electronically storing a plurality of parameter vector, each parameter vector being associated with a respective context and including one or more parameters, parameterizing the forecast model based on parameters provided in the first parameter vector to provide a parameterized forecast model, optimizing the parameterized forecast model to provide an optimized forecast model, and forecasting one or more values using the optimized forecast model.

Claims

1. A computer-implemented method for providing at least one parameter for use with a forecast model, the method being executed using one or more processors and comprising:

receiving a first context vector, the first context vector comprising a plurality of context attributes that describe a first context;

retrieving a first parameter vector from a repository based on the first context vector, the repository electronically storing a plurality of parameter vector, each parameter vector being associated with a respective context and comprising one or more parameters;

parameterizing the forecast model based on parameters provided in the first parameter vector to provide a parameterized forecast model;

optimizing the parameterized forecast model to provide an optimized forecast model; and

forecasting one or more values using the optimized forecast model.

2. The method of claim 1, wherein retrieving a first parameter vector from a repository comprises:

identifying a sub-plurality of parameter vectors based on the first context vector, the sub-plurality of parameter vectors including the first parameter vector;

for each parameter vector in the sub-plurality of parameter vectors, determining a cross-correlation coefficient; and

determining that a cross-correlation coefficient associated with the first parameter vector is the highest cross-correlation coefficient of the parameter vectors in the sub-plurality of parameter vectors and, in response, retrieving the first parameter vector.

3. The method of claim 2, wherein each cross-correlation coefficient reflects a distance between context attributes of the respective context and context attributes of the first context vector.

4. The method of claim 2, wherein each cross-correlation factor is provided as a Pearson cross-correlation coefficient.

5. The method of claim 1, wherein the repository is organized into a binary search tree.

6. The method of claim 5, wherein the binary search tree comprises a plurality of decision nodes and a plurality of leaf nodes, each decision node associated with a context attribute, each leaf node associated with one or more parameter vectors provided in a sub-plurality of parameter vectors.

7. The method of claim 6, wherein at least one decision node contains a splitting context attribute, a splitting value, and references to a left successor node and a right successor node.

8. The method of claim 5, wherein retrieving a first parameter vector comprises traversing nodes of the binary search tree based on context attributes of the first context vector.

9. The method of claim 8, further comprising:

re-traversing nodes of the binary search tree based on context attributes to retrieve a second parameter vector; and

parameterizing the forecast model based on parameters provided in the second parameter set to provide the parameterized forecast model.

10. The method of claim 9, wherein re-traversing nodes of the binary search tree comprises performing a ball-overlap-bounds test at a decision node to evaluate an existence of additional parameter vectors that may be present in an opposite branch of the decision node.

11. The method of claim 1, further comprising:

receiving a second context vector, the second context vector comprising a plurality of context attributes that describe a second context;

retrieving a second parameter vector from the repository based on the second context vector;

parameterizing the forecast model based on parameters provided in the second parameter vector to provide the parameterized forecast model;

optimizing the parameterized forecast model to provide the optimized forecast model; and

forecasting one or more values using the optimized forecast model.

12. The method of claim 11, wherein receiving a second context vector occurs in response to a context drift from the first context to the second context.

13. The method of claim 12, wherein the context drift comprises one of an abrupt drift, a persistent drift and a cyclic drift.

14. The method of claim 1, further comprising:

receiving the first parameter vector and a respective context vector; and

storing the first parameter vector in the repository based on context attributes of the respective context vector.

15. The method of claim 14, wherein storing the first parameter vector comprises:

traversing a binary search tree structure of the repository to identify a leaf node having a context that is most similar to the respective context vector; and

inserting the first parameter vector into the leaf node.

16. The method of claim 14, wherein storing the first parameter vector comprises:

traversing a binary search tree structure of the repository to identify a leaf node having a context that is most similar to the respective context vector; and

determining that a number of parameter vectors exceeds a maximum leaf node capacity and, in response;

adding at least two child nodes to the leaf node,

splitting parameter vectors of the leaf node into at least two groups,

populating that at least two child nodes with the at least two groups, respectively, and

converting the leaf node into a decision node with references to the at least two child nodes.

17. The method of claim 16, wherein splitting the parameter vectors into at least two groups is performed based on a splitting attribute and a split value that are determined from the models stored in the leaf node and based on a median value.

18. The method of claim 1, wherein context attributes comprise one or more of numerical attributes, nominal attributes and cyclic attributes.

19. A non-transitory computer-readable storage medium coupled to one or more processors and having instructions stored thereon which, when executed by the one or more processors, cause the one or more processors to perform operations for providing at least one parameter for use with a forecast model, the operations comprising:

receiving a first context vector, the first context vector comprising a plurality of context attributes that describe a first context;

retrieving a first parameter vector from a repository based on the first context vector, the repository electronically storing a plurality of parameter vector, each parameter vector being associated with a respective context and comprising one or more parameters;

parameterizing the forecast model based on parameters provided in the first parameter vector to provide a parameterized forecast model;

optimizing the parameterized forecast model to provide an optimized forecast model; and

forecasting one or more values using the optimized forecast model.

20. A system, comprising:

a computing device; and

a computer-readable storage device coupled to the computing device and having instructions stored thereon which, when executed by the computing device, cause the computing device to perform operations for providing at least one parameter for use with a forecast model, the operations comprising;

receiving a first context vector, the first context vector comprising a plurality of context attributes that describe a first context;

retrieving a first parameter vector from a repository based on the first context vector, the repository electronically storing a plurality of parameter vector, each parameter vector being associated with a respective context and comprising one or more parameters;

parameterizing the forecast model based on parameters provided in the first parameter vector to provide a parameterized forecast model;

optimizing the parameterized forecast model to provide an optimized forecast model; and

forecasting one or more values using the optimized forecast model.