Method and system for enhancing credit line management, price management and other discretionary levels setting for financial accounts

• US 8,073,767 B1
• Filed: 09/28/2009
• Issued: 12/06/2011
• Est. Priority Date: 05/21/2003
• Status: Active Grant

• Current Assignee: JP Morgan Chase Bank, N.A. (JP Morgan Chase & Co)
• Original Assignee: JP Morgan Chase Bank, N.A. (JP Morgan Chase & Co)
• Inventors: Pederson, Shane P., Lau, Tak Wing (Edward), Ma, Lizhi, Wang, Hui (Stephanie), Trench, Margaret S, Nair, Suresh K
• Primary Examiner(s): Cheung, Mary
• Application Number:
  US12/568,371
• Time in Patent Office: 799 Days
• Field of Search: 705 35- 44, 715/816
• US Class Current: 705/38
• CPC Class Codes: G06Q 20/10 specially adapted for electronic funds transfer [EFT] systems; specially adapted for home banking systems
• CPC Class Codes: G06Q 20/108 Remote banking, e.g. home banking
• CPC Class Codes: G06Q 20/403 Solvency checks
• CPC Class Codes: G06Q 40/00 Finance; Insurance; Tax strategies; Processing of corporate or income taxes
• CPC Class Codes: G06Q 40/02 Banking, e.g. interest calculation or account maintenance credit or loans G06Q40/03
• CPC Class Codes: G06Q 40/03 Credit; Loans; Processing thereof

Abstract

A method and system is disclosed for enabling the accurate determination of price points (APRs), credit limits, and other discretionary levels for each cardholder that maximize Net Present Value (NPV) for the portfolio, given constraints on quantities such as risk of default. In accordance with one embodiment, the present invention uses a Markov Decision Process (MDP) methodology to generate a simplified transition matrix representative of the potential state transitions for account holders. This model applies account level historical information on purchases, payments, profitability and delinquency risk to make these decisions. In addition, a data structure is disclosed constructed to implement a transition matrix computationally in different sizes.

Claims

1. A computer implemented method for managing financial accounts by changing at least one control variable including at least one of a credit line and an annual percentage rate, the method comprising:

collecting and storing in one or more storage devices account data for a plurality of financial accounts, where the account data are associated with a plurality of variables;

establishing at least one control variable and at least one behavior variable based on the plurality of variables;

determining, by a data processor, a plurality of interval breakpoints for the at least one control variable and the at least one behavior variable;

defining a plurality of account states based on the at least one control variable, the at least one behavior variable and the plurality of interval breakpoints;

forming a transition matrix based at least in part on the account data and the plurality of account states;

generating a Markov Decision Process (MDP) model based at least in part on the transition matrix;

generating a policy lookup table based on an output of the Markov Decision Process model, the policy lookup table listing pricing actions to be taken in response to the plurality of account states, each pricing action comprising changing or maintaining at least one control variable including at least one of a credit line and an annual percentage rate;

determining behavior variable data associated with a financial account;

slotting the financial account into one of the plurality of account states based at least in part on the determined behavior variable data associated with the financial account; and

determining, from the policy lookup table and based on the slotted account state, at least one pricing action to be taken on the financial account.

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

applying business rules to the output of the Markov Decision Process model.

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

periodically re-evaluating pricing actions for at least one of the plurality of financial accounts by;

determining behavior variable data associated with the at least one financial account;

slotting the at least one financial account into one of the plurality of account states based at least in part on the determined behavior variable data associated with the at least one financial account; and

determining, from the policy lookup table and based on the slotted account state, at least one pricing action to be taken on the at least one financial account.

4. The method according to claim 1, where the account data comprise time-series data on at least one of:

(a) credit bureau variables;

(b) credit bureau scores; and

(c) bankcard account performance.

5. The method according to claim 1, where the policy lookup table is generated to achieve a desired effect associated with at least one of the financial accounts'"'"':

(a) net cash flow (NCF);

(b) balance migration;

or

(c) risk of credit losses.

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

performing a regression tree analysis on the plurality of variables;

establishing the at least one behavior variable based on the regression tree analysis; and

determining the plurality of interval breakpoints for each of the at least one behavior variable based on the regression tree analysis.

7. The method according to claim 1, further comprising simplifying the transition matrix based on one or more of the following techniques:

(a) creating a rectangular transition matrix with rows for the at least one control variable and the at least one behavioral variable, and columns for the at least one behavior variable only;

(b) combining at least one row of a transition matrix with an adjoining row, where the at least one row is characterized with low-frequency transitions;

(c) combining at least one column of a transition matrix with an adjoining column, where the at least one column is characterized with low-frequency transitions; and

(d) re-indexing a transition matrix to correctly capture all past and future account transitions.

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

validating and supplementing the output of the Markov Decision Process model through a Monte Carlo simulation, where the Monte Carlo simulation is based at least partially on the account data.

9. The method of claim 1, wherein a Portfolio Control and Optimization model is represented as:

$V_t<br><br>\left(x,y\right)=\{\begin{array}{cc}\begin{array}{c}\underset{a,\in <br><br>A<br><br>\left(x,y\right)}{\max }<br><br>r<br><br>\left(x\pm <br><br>a_t,y\right)+\\ \mathrm{\beta <br><br>}<br><br>\underset{j\in <br><br>S}{\sum <br><br>}<br><br><br><br>p<br><br>\left(x\pm <br><br>a_t,y;<br><br>j\right)<br><br>V_{t+1}<br><br>\left(x\pm <br><br>a_t,j\right)\end{array}& \mathrm{if}<br><br><br><br>t=\mathrm{update}<br><br><br><br>\mathrm{epoch}\\ r<br><br>\left(x,y\right)+\mathrm{\beta <br><br>}<br><br>\underset{j\in <br><br>S}{\sum <br><br>}<br><br>p<br><br>\left(x,y;<br><br>j\right)<br><br>V_{t+1}<br><br>\left(x,j\right)& \mathrm{otherise}\end{array}$
V_T(x,y)=r(x,y)

wherein,

V_t(x,y) is the optimal discounted Net Present Value (NPV) in state (x,y) and time t;

A(x,y) is a set of actions α

_t available in state (x,y);

r(x,y) is the Net Cash Flow (NCF) in state (x,y);

r(x±

α

_t, y) is the Net Cash Flow (NCF) in state (x,y) when offset for the action α

_t taken;

β

is a one period discount factor;

p(x±

α

_t,y;

j) is the transition probability that specifies the likelihood of moving to state j from the sate (x±

α

_t,y);

S is the set of all states; and

T is the time horizon for which the model is solved.

10. A non-transitory computer readable medium for causing a processor to manage financial accounts by changing at least one control variable including at least one of a credit line and an annual percentage rate, the computer readable medium comprising one or more instructions for performing the following:

collecting and storing in one or more storage devices account data for a plurality of financial accounts, where the account data are associated with a plurality of variables;

establishing at least one control variable and at least one behavior variable based on the plurality of variables;

determining, by a data processor, a plurality of interval breakpoints for the at least one control variable and the at least one behavior variable;

defining a plurality of account states based on the at least one control variable, the at least one behavior variable and the plurality of interval breakpoints;

forming a transition matrix based at least in part on the account data and the plurality of account states;

generating a Markov Decision Process (MDP) model based at least in part on the transition matrix;

generating a policy lookup table based on an output of the Markov Decision Process model, the policy lookup table listing pricing actions to be taken in response to the plurality of account states, each pricing action comprising changing or maintaining at least one control variable including at least one of a credit line and an annual percentage rate;

determining behavior variable data associated with a financial account;

slotting the financial account into one of the plurality of account states based at least in part on the determined behavior variable data associated with the financial account; and

determining, from the policy lookup table and based on the slotted account state, at least one pricing action to be taken on the financial account.

11. The non-transitory computer readable medium according to claim 10, further comprising instructions for:

applying business rules to the output of the Markov Decision Process model.

12. The non-transitory computer readable medium according to claim 10, further comprising instructions for:

periodically re-evaluating pricing actions for at least one of the plurality of financial accounts by;

determining behavior variable data associated with the at least one financial account;

slotting the at least one financial account into one of the plurality of account states based at least in part on the determined behavior variable data associated with the at least one financial account; and

determining, from the policy lookup table and based on the slotted account state, at least one pricing action to be taken on the at least one financial account.

13. The non-transitory computer readable medium according to claim 10, where the account data comprise time-series data on at least one of:

(a) credit bureau variables;

(b) credit bureau scores; and

(c) bankcard account performance.

14. The non-transitory computer readable medium according to claim 10, where the policy lookup table is generated to achieve a desired effect associated with at least one of the financial accounts'"'"':

(a) net cash flow (NCF);

(b) balance migration;

or

(c) risk of credit losses.

15. The non-transitory computer readable medium according to claim 10, further comprising instructions for:

performing a regression tree analysis on the plurality of variables;

establishing the at least one behavior variable based on the regression tree analysis; and

determining the plurality of interval breakpoints for each of the at least one behavior variable based on the regression tree analysis.

16. The non-transitory computer readable medium according to claim 10, further comprising instructions for simplifying the transition matrix based on one or more of the following techniques:

(a) creating a rectangular transition matrix with rows for the at least one control variable and the at least one behavioral variable, and columns for the at least one behavior variable only;

(b) combining at least one row of a transition matrix with an adjoining row, where the at least one row is characterized with low-frequency transitions;

(c) combining at least one column of a transition matrix with an adjoining column, where the at least one column is characterized with low-frequency transitions; and

(d) re-indexing a transition matrix to correctly capture all past and future account transitions.

17. The non-transitory computer readable medium according to claim 10, further comprising instructions for validating and supplementing the output of the Markov Decision Process model through a Monte Carlo simulation, where the Monte Carlo simulation is based at least partially on the account data.

18. A computer implemented system for managing financial accounts by changing at least one control variable including at least one of a credit line and an annual percentage rate, the method, the system comprising:

at least one data processor and one or more storage devices that are adapted to;

collect and store in the one or more storage devices account data for a plurality of financial accounts, where the account data are associated with a plurality of variables;

establish at least one control variable and at least one behavior variable based on the plurality of variables;

determine a plurality of interval breakpoints for the at least one control variable and the at least one behavior variable;

define a plurality of account states based on the at least one control variable, the at least one behavior variable and the plurality of interval breakpoints;

form a transition matrix based at least in part on the account data and the plurality of account states;

generate a Markov Decision Process (MDP) model based at least in part on the transition matrix;

generate a policy lookup table based on an output of the Markov Decision Process model, the policy lookup table listing pricing actions to be taken in response to the plurality of account states, each pricing action comprising changing or maintaining at least one control variable including at least one of a credit line and an annual percentage rate;

determine behavior variable data associated with a financial account;

slot the financial account into one of the plurality of account states based at least in part on the determined behavior variable data associated with the financial account; and

determine, from the policy lookup table and based on the slotted account state, at least one pricing action to be taken on the financial account.

19. The system according to claim 18, wherein the at least one processor is further adapted to:

apply business rules to the output of the Markov Decision Process model.

20. The system according to claim 18, wherein the at least one processor is further adapted to:

periodically re-evaluate pricing actions for at least one of the plurality of financial accounts by;

determining behavior variable data associated with the at least one financial account;

slotting the at least one financial account into one of the plurality of account states based at least in part on the determined behavior variable data associated with the at least one financial account; and

determining, from the policy lookup table and based on the slotted account state, at least one pricing action to be taken on the at least one financial account.

21. The system according to claim 18, where the account data comprise time-series data on at least one of:

(a) credit bureau variables;

(b) credit bureau scores; and

(c) bankcard account performance.

22. The system according to claim 18, where the policy lookup table is generated to achieve a desired effect associated with at least one of the financial accounts'"'"':

(a) net cash flow (NCF);

(b) balance migration;

or

(c) risk of credit losses.

23. The system according to claim 18, wherein the at least one processor is further adapted to:

perform a regression tree analysis on the plurality of variables;

establish the at least one behavior variable based on the regression tree analysis; and

determine the plurality of interval breakpoints for each of the at least one behavior variable based on the regression tree analysis.

24. The system according to claim 18, wherein the at least one processor is further adapted to simplify the transition matrix based on one or more of the following techniques:

(a) creating a rectangular transition matrix with rows for the at least one control variable and the at least one behavioral variable, and columns for the at least one behavior variable only;

(b) combining at least one row of a transition matrix with an adjoining row, where the at least one row is characterized with low-frequency transitions;

(c) combining at least one column of a transition matrix with an adjoining column, where the at least one column is characterized with low-frequency transitions; and

(d) re-indexing a transition matrix to correctly capture all past and future account transitions.

25. The system according to claim 18, wherein the at least one processor is further adapted to:

validate and supplement the output of the Markov Decision Process model through a Monte Carlo simulation, where the Monte Carlo simulation is based at least partially on the account data.