Scorecard models with measured variable interactions

• US 9,367,520 B2
• Filed: 12/20/2012
• Issued: 06/14/2016
• Est. Priority Date: 12/20/2012
• Status: Active Grant

• Current Assignee: Fair Isaac Corporation
• Original Assignee: Fair Isaac Corporation
• Inventors: Zhao, Xing, Hamilton, Peter, Story, Andrew K., Flint, Andrew
• Primary Examiner(s): Rivas, Omar Fernandez
• Assistant Examiner(s): Khan, Iftekhar
• Application Number:
  US13/722,826
• Publication Number:
  US 20140180649A1
• Time in Patent Office: 1,272 Days
• US Class Current: 1/1
• 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

Abstract

Data is received that characterizes a transaction and includes a plurality of values corresponding to variables. Thereafter, a score is determined for the transaction based on the received data and using a scoring model. The scoring model only uses variables pairs having a divergence residual above a pre-defined threshold. Thereafter, data is provided that characterizes the determined score. Related apparatus, systems, techniques and computer program products are also described.

Claims

1. A computer-implemented method comprising:

receiving data characterizing a transaction, the data comprising a plurality of values corresponding to variables;

determining, based on the received data and using a scoring model, a score for the transaction, the scoring model only using variables pairs having a divergence residual above a pre-defined threshold, the divergence residual being a difference between a divergence of an N-dimensional scorecard model and a divergence of an N-variable scorecard for a pair of variables, wherein N is an integer greater than or equal to 2; and

providing data characterizing the determined score;

wherein at least one of the receiving, the determining, and the providing is performed by at least one processor of at least one computing system.

2. A method as in claim 1, wherein providing data comprises one or more of:

displaying the score, transmitting data characterizing the determined score, loading data characterizing the determined score, and storing data characterizing the determined score.

3. A method as in claim 1, wherein the scoring model is a scorecard model.

4. A method as in claim 3, wherein the scoring model is a two dimensional (2-D) scorecard model.

5. A method as in claim 3, wherein the scoring model is an N dimensional scorecard model, wherein N is greater than 2.

6. A method as in claim 3, wherein the scorecard model is developed by capturing interactions among cross binnings between a given predictor pair (x_i1, x_i2).

7. A method as in claim 6, wherein the interactions among cross binnings are captured using:

8. A method as in claim 7, wherein the divergence residual, DivR (x_i1, x_i2), is equal to Div(x_i1×

x_i2)−

Div(x_i1+x_i2), wherein Div(x_i1×

x_i2) is divergence of an N dimensional scorecard model, and Div(x_i1+x_i2) is divergence of an N-variable scorecard.

9. A method as in claim 1, wherein the scoring model ƒ

comprises;

10. A computer-implemented method comprising:

receiving data characterizing a transaction, the data comprising a plurality of values corresponding to variables;

determining, based on the received data and using a scoring model, a score for the transaction, the scoring model only using variables pairs having a root mean square error (RMSE) residual above a pre-defined threshold, the RMSE residual being a difference between a RMSE of an N-dimensional scorecard model and a RMSE of an N-variable scorecard for a pair of variables, wherein N is an integer greater than or equal to 2; and

providing data characterizing the determined score;

wherein at least one of the receiving, the determining, and the providing is performed by at least one processor of at least one computing system.

11. A method as in claim 10, wherein providing data comprises one or more of:

displaying the score, transmitting data characterizing the determined score, loading data characterizing the determined score, and storing data characterizing the determined score.

12. A method as in claim 10, wherein the scoring model is a two dimensional (2-D) scorecard model.

13. A method as in claim 12, wherein the scorecard model is developed by capturing interactions among cross binnings between a given predictor pair (x_i1, x_i2).

14. A method as in claim 13, wherein the interactions among cross binnings are captured using:

15. A computer-implemented method comprising:

receiving data comprising a plurality of values for variables across a plurality of transactions;

identifying pairs of variables having divergence values above a pre-defined threshold, the identifying being determined based on a divergence residual being a difference between a divergence of an N-dimensional scorecard model and a divergence of an N-variable scorecard for a pair of variables, wherein N is an integer greater than or equal to 2; and

generating a scorecard model using the identified pairs of variables;

wherein at least one of the receiving, the identifying, and the generating is performed by at least one processor of at least one computing system.

16. A method as in claim 15, wherein the scorecard model is a segmented scorecard model and the identified pairs of variables are used as splitter variables.

17. A computer-implemented method comprising:

receiving data comprising a plurality of values for variables across a plurality of transactions;

identifying pairs of variables having RMSE residual values above a pre-defined threshold, the RMSE residual being a difference between a RMSE of an N-dimensional scorecard model and a RMSE of an N-variable scorecard for a pair of variables, wherein N is an integer greater than or equal to 2; and

generating a scorecard model using the identified pairs of variables;

wherein at least one of the receiving, the identifying, and the generating is performed by at least one processor of at least one computing system.

18. A method as in claim 17, wherein the scorecard model is a segmented scorecard model.

19. A method as in claim 18, wherein the segmented scorecard model is found using a segmented scorecard search procedure.

20. A method as in claim 19, wherein the identified pairs of variables are used as splitter variables.