Apparatus and method for simulating an analytic value chain

• US 7,702,576 B2
• Filed: 06/20/2008
• Issued: 04/20/2010
• Est. Priority Date: 08/09/2005
• Status: Active Grant

• Current Assignee: Fair Isaac Corporation
• Original Assignee: Fair Isaac Corporation
• Inventors: Milana, Joseph P., Fahner, Gerald
• Primary Examiner(s): Borlinghaus; Jason M
• Application Number:
  US12/143,428
• Publication Number:
  US 20080313073A1
• Time in Patent Office: 669 Days
• Field of Search: None
• US Class Current: 705/38
• CPC Class Codes: G06Q 10/00 Administration; Management
• CPC Class Codes: G06Q 20/10 specially adapted for electronic funds transfer [EFT] systems; specially adapted for home banking systems
• CPC Class Codes: G06Q 40/00 Finance; Insurance; Tax strategies; Processing of corporate or income taxes
• CPC Class Codes: G06Q 40/03 Credit; Loans; Processing thereof

Abstract

A computer-implemented simulator models the entire analytic value chain so that data generation, model fitting and strategy optimization are an integral part of the simulation. Data collection efforts, data mining algorithms, predictive modeling technologies and strategy development methodologies define the analytic value chain of a business operation: data→models→strategies→profit. Inputs to the simulator include consumer data and potential actions to be taken regarding a consumer or account. The invention maps what is known about a consumer or an account and the potential actions that the business can take on that consumer or account to potential future financial performance. After iteratively performing simulations using varying inputs, modeling the effect of the innovation on a profit model, the simulator outputs a prediction of the commercial value of an analytic innovation.

Claims

1. A computer-implemented method including a learning strategy for simulating future outcomes of a credit line increase decision strategy, the method being implemented by one or more data processors and comprising:

providing, by at least one of the data processors, an initial model for a time-dependent relationship for expected profit from a credit card operation that is controlled by a line increase decision strategy;

iteratively simulating the future outcomes wherein the simulating is jump-started with an empirical sample and wherein associated actions are then simulated according to a first posited learning strategy;

calculating, by at least one of the data processors, mean profit per account from the first posited learning strategy using the initial model;

simulating, by at least one of the data processors, future outcomes based on the parameters of the initial profit model wherein an initial strategy development set is generated;

allowing, by at least one of the data processors, the profit model to vary over time, as indicated by a time variant parameter vector;

over learning cycles t=1, . . . , T;

estimating, by at least one of the data processors, a profit model based on a previous strategy development set;

estimating, by at least one of the data processors, an optimal strategy resulting in recommended line increases for the learning strategy;

stochastically assigning, by at least one of the data processors, test actions according to a test design;

calculating, by at least one of the data processors, mean profit per account for the learning strategy from a posited model; and

updating, by at least one of the data processors, the development data set by simulating future account behavior from a parameter vector from the posited model, where t⇄

t+1;

wherein the learning strategy comprises a decision tree, wherein leaves of the decision tree define a plurality of segments, wherein a segment is associated with a recommended line increase and a test design for an alternative increase level, wherein, for an account falling into a segment, a simulated random line increase is assigned to an amount specified for the segment according to a multinomial distribution.

2. A method as in claim 1, wherein the relationship stays constant over a cycle and jumps between cycles.

3. A method as in claim 2, wherein the cycles are discrete.

4. A method as in claim 1, further comprising:

positing, by at least one of the data processors, certain parameters for error distributions for the initial simulation.

5. A method as in claim 1, further comprising:

performing, by at least one of the data processors, a dynamic simulation for learning strategies that differ in their range of testing.

6. A method as in claim 5, further comprising:

providing, by at least one of the data processors, a conservative learning strategy that tests in a narrower range than other ranges around a recommended increase, and the probabilities that tests are taken are small, resulting in a small number of tests overall as compared to other ranges.

7. A method as in claim 5, further comprising:

providing, by at least one of the data processors, a strategy having a range greater than other strategy ranges, wherein the learning strategy performs testing in term of test ranges and test frequencies, resulting in a large number of tests overall as compared to other ranges.

8. A method as in claim 5, further comprising:

keeping, by at least one of the data processors, the profit model constant for a plurality of learning cycles and at a predetermined iteration changing the profit model and leaving it constant thereafter.

9. A computer program product comprising a tangible machine-readable storage medium embodying instructions that when performed by one or more machines result in operations comprising:

providing an initial model for a time-dependent relationship for expected profit from a credit card operation that is controlled by a line increase decision strategy;

iteratively simulating the future outcomes wherein the simulating is initiated with an empirical sample and wherein associated actions are then simulated according to a first posited learning strategy;

calculating mean profit per account from the first posited learning strategy using the initial model;

simulating future outcomes based on the parameters of the initial profit model wherein an initial strategy development set is generated, the simulating positing certain parameters for error distributions for the initial simulation;

allowing the profit model to vary over time, as indicated by a time variant parameter vector;

over learning cycles t=1, . . . , T;

estimating a profit model based on a previous strategy development set;

estimating an optimal strategy resulting in recommended line increases for the learning strategy;

stochastically assigning test actions according to a test design;

calculating mean profit per account for the learning strategy from a posited model;

updating the development data set by simulating future account behavior from a parameter vector from the posited model, where

wherein the learning strategy comprises a decision tree, wherein leaves of the decision tree define a plurality of segments, wherein a segment is associated with a recommended line increase and a test design for an alternative increase level, wherein, for an account falling into a segment, a simulated random line increase is assigned to an amount specified for the segment according to a multinomial distribution.

10. A computer program product as in claim 9, wherein the relationship stays constant over a cycle and jumps between cycles.

11. A computer program product as in claim 10, wherein the cycles are discrete.

12. A computer program product as in claim 9, wherein the learning strategy comprises a decision tree, wherein leaves of the decision tree define a plurality of segments, wherein a segment is associated with a recommended line increase and a test design for an alternative increase level, wherein, for an account falling into a segment, a simulated random line increase is assigned to an amount specified for the segment according to a multinomial distribution.

13. A computer program product as in claim 9, wherein the tangible machine-readable storage medium further embodies instructions that when performed by one or more machines result in operations comprising:

performing a dynamic simulation for learning strategies that differ in their range of testing.

14. A computer program product as in claim 13, wherein the tangible machine-readable storage medium further embodies instructions that when performed by one or more machines result in operations comprising:

providing a conservative learning strategy that tests in a narrower range than other ranges around a recommended increase, and the probabilities that tests are taken are small, resulting in a small number of tests overall as compared to other ranges.

15. A computer program product as in claim 13, wherein the tangible machine-readable storage medium further embodies instructions that when performed by one or more machines result in operations comprising:

providing a strategy having a range greater than other strategy ranges, wherein the learning strategy performs testing in term of test ranges and test frequencies, resulting in a large number of tests overall as compared to other ranges.

16. A computer program product as in claim 13, wherein the tangible machine-readable storage medium further embodies instructions that when performed by one or more machines result in operations comprising:

keeping the profit model constant for a plurality of learning cycles and at a predetermined iteration changing the profit model and leaving it constant thereafter.