Methods and systems for analytical-based multifactor multiobjective portfolio risk optimization

• US 7,593,880 B2
• Filed: 03/19/2003
• Issued: 09/22/2009
• Est. Priority Date: 03/19/2003
• Status: Expired due to Fees

• Current Assignee: GE Financial Assurance Holdings Incorporated (General Electric Company), General Electric Company
• Original Assignee: General Electric Company
• Inventors: Kiaer, Carol Lynn, Bollapragada, Srinivas, Chalermkraivuth, Kete Charles, Chakraborty, Anindya
• Primary Examiner(s): Havan; Thu Thao
• Assistant Examiner(s): Hammond, III; Thomas M
• Application Number:
  US10/390,709
• Publication Number:
  US 20040186814A1
• Time in Patent Office: 2,379 Days
• Field of Search: 705 35- 40
• US Class Current: 705/36.00R
• CPC Class Codes: G06Q 40/06 Asset management; Financial planning or analysis
• CPC Class Codes: G06Q 40/08 Insurance

Abstract

The invention provides systems and methods for determining an efficient frontier, which comprises a collection of security allocations in a portfolio, with multiple, conflicting objectives in a multi-factor portfolio problem. The method includes providing a mathematical model of a relaxation of a problem; generating a sequence of additional constraints; and sequentially applying respective nonlinear risk functions to generate respective adjusted maximum return solutions to obtain an efficient frontier.

Claims

1. A method for determining optimal solutions along an efficient frontier comprising, a collection of security allocations for a portfolio by using a sequential linear programming technique for a problem having a nonlinear constraint, the method comprising:

(a) generating a first optimal solution point along the efficient frontier by solving a relaxed linear programming problem with a plurality of linear constraints that excludes the nonlinear constraint using a processor of a computing device, the plurality of linear constraints forming a feasible region;

(b) calculating a first tangent plane for the nonlinear constraint at said first optimal solution point using the processor of the computing device;

(c) shifting the first tangent plane for the nonlinear constraint by a small step size to produce an additional linear constraint using the processor of the computing device, wherein the additional linear constraint lies within the feasible region;

(d) generating a second optimal solution point along the efficient frontier by solving the relaxed linear programming problem with the additional linear constraint using the processor of the computing device;

(e) calculating a second tangent plane for the nonlinear constraint at said second optimal solution point using the processor of the computing device;

(f) shifting the second tangent plane for the nonlinear constraint by a small step size to produce a second additional linear constraint using the processor of the computing device, wherein the second additional linear constraint lies within the feasible region; and

(g) generating additional optimal solution points along the efficient frontier by solving the relaxed linear programming problem with additional linear constraints using the processor of the computing device.

2. The method of claim 1, further comprising the step of applying nonlinear risk functions at each optimum solution point to generate a respective adjusted maximum return solutions.

3. The method of claim 2, wherein applying nonlinear risk functions to generate respective adjusted maximum return solutions further includes:

evaluating a nonlinear risk fUnction at the maximum return solution to determine the risk at that point, which is defined by the maximum return solution;

determining a tangent plane of that nonlinear risk function at the maximum return solution;

determining a step size based on the desired granularity of the efficient frontier;

shifting the tangent plane by the determined step size to provide a first shifted tangent plane;

applying the first shifted tangent plane as a new linear constraint; and

generating an adjusted feasible region and corresponding maximum return solution.

4. The method of claim 1, further including performing a test for local convexity of nonlinear risk functions.

5. The method of claim 3, wherein the step size varies with a current adjusted maximum return solution.

6. The method of claim 5, further including determining a step change, between successive steps.

7. The method of claim 5, wherein the step size is based on a calculation of a distance between a current nonlinear function value, being a risk level, and the target value, being a minimum risk value.

8. The method of claim 7, wherein a preliminary step size is given by the distance divided by the desired number of steps, and the preliminary step size is adjusted with information obtained by testing a terrain around a current solution.

9. The method of claim 1, wherein the portfolio is a collection of securities including at least one of assets and liabilities.

19. The method of claim 1, wherein steps (b)-(g) are repeated for a plurality of nonlinear constraints.

10. An article of manufacture for determining an efficient frontier comprising a collection of security allocations in a portfolio by using a sequential linear programming technique for a problem having a nonlinear constraint, the article of manufacture comprising:

a computer readable medium tangibly embodying a computer program executable by a processor of a computer, wherein the computer program contains a model portion that generates a first optimal solution point along the efficient frontier by solving a relaxed linear programming problem with a plurality of linear constraints that form a feasible region, wherein the solving excludes the nonlinear constraint, calculates a first tangent plane for the nonlinear constraint at said first optimal solution point, shifts the first tangent plane for the nonlinear constraint by a small step size to produce an additional linear constraint that lies within the feasible region, and generates a second optimal solution point along the efficient frontier by solving the relaxed linear programming problem with the additional linear constraint;

a linear constraint generation portion that generates a sequence of additional linear constraints, wherein said additional linear constraints are an approximation of the nonlinear constraint; and

a solution generation portion that sequentially applies the sequence of additional linear constraints to generate additional optimal solution points along the efficient frontier.

11. The article of manufacture of claim 10, wherein the solution generation portion sequentially applies respective nonlinear risk functions to generate respective adjusted maximum return solutions.

12. The article of manufacture of claim 11, wherein the solution generation portion further includes:

evaluating a nonlinear risk function at the maximum return solution to determine the risk at that point, which is defined by the maximum return solution;

using a gradient of the nonlinear risk function to determine a tangent plane of that nonlinear risk function at the maximum return solution;

determining a step size based on the desired granularity of the efficient frontier;

shifting the tangent plane by a determined step size to provide a shifted tangent plane;

applying the first shifted tangent plane as a new linear constraint; and

generating an adjusted feasible region and corresponding maximum return solution.

13. The article of manufacture of claim 10, wherein the solution generation portion performs a test for local convexity of a nonlinear risk function.

14. The article of manufacture of claim 13, wherein the step size varies with a current adjusted maximum return solution.

15. The article of manufacture of claim 14, wherein the solution generation portion determines a step change, between successive steps.

16. The article of manufacture of claim 10, wherein the portfolio is a collection of securities including at least one of assets and liabilities.

17. An article of manufacture readable by a machine, tangibly embodying a program of instructions executable by the machine to perform method steps for determining an efficient frontier, comprising a collection of security allocations in a portfolio, said method steps comprising:

(a) generating a first optimal solution point along the efficient frontier by solving a relaxed linear programming problem with a plurality of linear constraints that excludes the nonlinear constraint using a processor of a computing device, the plurality of linear constraints forming a feasible region;

(b) calculating a first tangent plane for the nonlinear constraint at said first optimal solution point using the processor of the computing device;

(c) shifting the first tangent plane for the nonlinear constraint by a small step size to produce an additional linear constraint using the processor of the computing device, wherein the additional linear constraint lies within the feasible region;

(d) generating a second optimal solution point along the efficient frontier by solving the relaxed linear programming problem with the additional linear constraint using the processor of the computing device;

(e) calculating a second tangent plane for the nonlinear constraint at said second optimal solution point using the processor of the computing device;

(f) shifting the second tangent plane for the nonlinear constraint by a small step size to produce a second additional linear constraint using the processor of the computing device, wherein the second additional linear constraint lies within the feasible region; and

(g) generating additional optimal solution points along the efficient frontier by solving the relaxed linear programming problem with additional linear constraints using the processor of the computing device.

18. The article of manufacture of claim 17, further comprising the steps of:

evaluating a nonlinear risk function at the maximum return solution to determine the risk at that point, which is defined by the maximum return solution;

using a gradient of the nonlinear risk function to determine a tangent plane of that nonlinear risk function at the maximum return solution;

determining a step size based on the desired granularity of the efficient frontier;

shifting the tangent plane by a determined step size to provide a shifted tangent plane;

applying the first shifted tangent plane as a new linear constraint; and

generating an adjusted feasible region and corresponding maximum return solution.

20. The article of manufacture of claim 17, wherein steps (b)-(g) are repeated for a plurality of nonlinear constraints.