Systems and methods for multi-objective portfolio analysis using pareto sorting evolutionary algorithms

• US 8,219,477 B2
• Filed: 02/20/2004
• Issued: 07/10/2012
• Est. Priority Date: 02/20/2004
• Status: Active Grant

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Subbu, Rajesh Venkat, Bollapragada, Srinivas, Bonissone, Piero Patrone, Chalermkraivuth, Kete Charles, Eklund, Neil Holger White, Iyer, Naresh Sundaram
• Primary Examiner(s): Borlinghaus, Jason M
• Application Number:
  US10/781,805
• Publication Number:
  US 20050187846A1
• Time in Patent Office: 3,063 Days
• Field of Search: 705/36.R
• US Class Current: 705/36.R
• CPC Class Codes: G06Q 40/06 Asset management; Financial planning or analysis

Abstract

The systems and methods of the invention are directed to portfolio optimization and related techniques. For example, the invention provides a method for multi-objective portfolio optimization for use in investment decisions based on competing objectives and a plurality of constraints constituting a portfolio problem, the method comprising: generating an initial population of solutions of portfolio allocations; committing the initial population of solutions to an initial population archive; performing a multi-objective process, based on the initial population archive and on multiple competing objectives, to generate an efficient frontier, the multi-objective process including a evolutionary algorithm process, the evolutionary algorithm process utilizing a dominance filter, the efficient frontier being used in investment decisioning.

Claims

1. A method for multi-objective portfolio analysis using Pareto Sorting Evolutionary Algorithms, the method comprising the steps of:

(a) randomly drawing an initial population of individual portfolio allocations that are generated from a portfolio allocations archive by using a combination of linear programming and sequential linear programming algorithms using a processor of a computing device;

(b) passing the initial population of portfolio allocations through a dominance filter to identify a non-dominated subset of parent portfolio allocations using the processor of the computing device;

(c) committing the non-dominated subset of parent portfolio allocations to a non-dominated portfolio allocations archive using the processor of the computing device;

(d) randomly combining matched pairs of parent portfolio allocations to create offspring portfolio allocations using the processor of the computing device;

(e) passing the offspring portfolio allocations through the dominance filter to identify a non-dominated subset of offspring portfolio allocations using the processor of the computing device;

(f) combining the non-dominated subset of parent portfolio allocations with the non-dominated subset of offspring portfolio allocations into a larger set of portfolio allocations using the processor of the computing device;

(g) passing the larger set of portfolio allocations through a non-crowding filter to identify a reduced subset of portfolio allocations using the processor of the computing device;

(h) creating a new population of individual portfolio allocations from the reduced subset of portfolio allocations using the processor of the computing device;

(i) updating the non-dominated portfolio allocations archive with the new population of individual portfolio allocations using the processor of the computing device;

(j) repeating steps (a) through (i) for a plurality of generations using the processor of the computing device; and

(k) passing the updated non-dominated portfolio allocations archive through the dominance filter to generate an interim efficient frontier in a portfolio performance space having at least three-dimensions using the processor of the computing device, the interim efficient frontier being used in investment decisions.

2. The method of claim 1, wherein the non-dominated subset of parent portfolio allocations has a first cardinality.

3. The method of claim 2, wherein the non-dominated subset of offspring portfolio allocations has a second cardinality that is different than the first cardinality.

4. The method of claim 3, wherein the larger set of portfolio allocations has a third cardinality that is equal to the first cardinality plus the second cardinality.

5. The method of claim 4, wherein the reduced subset of portfolio allocations has a fourth cardinality that is less than the third cardinality.

6. The method of claim 1, wherein steps (a)-(j) are repeated until convergence is achieved or allocated computational cycles are exhausted.

7. The method of claim 1, wherein the reduced subset of portfolio allocations has a fourth cardinality, and wherein the new population of individual portfolio allocations is exactly equal to the reduced set of portfolio allocations if the fourth cardinality is equal to the initial population of individual portfolio allocations.

8. The method of claim 1, wherein the reduced subset of portfolio allocations has a fourth cardinality, and wherein the new population of individual portfolio allocations is created by randomly drawing additional individual portfolio allocations from the portfolio allocations archive if the fourth cardinality is less than to the initial population of individual portfolio allocations.

9. The method of claim 1, wherein the reduced subset of portfolio allocations has a fourth cardinality, and wherein the new population of individual portfolio allocations is created by randomly discarding individual portfolio allocations from the reduced subset of portfolio allocations if the fourth cardinality is greater than to the initial population of individual portfolio allocations.

10. The method of claim 9, wherein the new population of individual portfolio allocations is created by randomly injecting individual portfolio allocations from the portfolio allocations archive until the fourth cardinality is equal to a desired number of individual portfolio allocations.

11. A system for multi-objective portfolio analysis using Pareto Sorting Evolutionary Algorithms comprising an efficient frontier processing portion that randomly draws an initial population of individual portfolio allocations that are generated from a portfolio allocations archive by using a combination of linear programming and sequential linear programming algorithms;

passes the initial population of portfolio allocations through a dominance filter to identify a non-dominated subset of parent portfolio allocations;

commits the non-dominated subset of parent portfolio allocations to a non-dominated portfolio allocations archive;

randomly combines matched pairs of parent portfolio allocations to create offspring portfolio allocations;

passes the offspring portfolio allocations through the dominance filter to identify a non-dominated subset of offspring portfolio allocations;

combines the non-dominated subset of parent portfolio allocations with the non-dominated subset of offspring portfolio allocations into a larger set of portfolio allocations;

passes the larger set of portfolio allocations through a non-crowding filter to identify a reduced subset of portfolio allocations;

creates a new population of individual portfolio allocations from the reduced subset of portfolio allocations;

updates the non-dominated portfolio allocations archive with the new population of individual portfolio allocations; and

passes the updated non-dominated portfolio allocations archive through the dominance filter to generate an interim efficient frontier in a portfolio performance space having at least three-dimensions, the interim efficient frontier being used in investment decisions.

12. The system of claim 11, wherein the non-dominated subset of parent portfolio allocations has a first cardinality.

13. The system of claim 12, wherein the non-dominated subset of offspring portfolio allocations has a second cardinality that is different than the first cardinality.

14. The system of claim 13, wherein the larger set of portfolio allocations has a third cardinality that is equal to the first cardinality plus the second cardinality.

15. The system of claim 14, wherein the reduced subset of portfolio allocations has a fourth cardinality that is less than the third cardinality.

16. The system of claim 11, wherein the reduced subset of portfolio allocations has a fourth cardinality, and wherein the new population of individual portfolio allocations is exactly equal to the reduced set of portfolio allocations if the fourth cardinality is equal to the initial population of individual portfolio allocations.

17. The system of claim 11, wherein the reduced subset of portfolio allocations has a fourth cardinality, and wherein the new population of individual portfolio allocations is created by randomly drawing additional individual portfolio allocations from the portfolio allocations archive if the fourth cardinality is less than to the initial population of individual portfolio allocations.

18. The system of claim 11, wherein the reduced subset of portfolio allocations has a fourth cardinality, and wherein the new population of individual portfolio allocations is created by randomly discarding individual portfolio allocations from the reduced subset of portfolio allocations if the fourth cardinality is greater than to the initial population of individual portfolio allocations.

19. The system of claim 18, wherein the new population of individual portfolio allocations is created by randomly injecting individual portfolio allocations from the portfolio allocations archive until the fourth cardinality is equal to a desired number of individual portfolio allocations.

20. A computer readable medium for multi-objective portfolio analysis using Pareto Sorting Evolutionary Algorithms comprising an efficient frontier processing portion that randomly draws an initial population of individual portfolio allocations that are generated from a portfolio allocations archive by using a combination of linear programming and sequential linear programming algorithms;

passes the initial population of portfolio allocations through a dominance filter to identify a non-dominated subset of parent portfolio allocations;

commits the non-dominated subset of parent portfolio allocations to a non-dominated portfolio allocations archive;

randomly combines matched pairs of parent portfolio allocations to create offspring portfolio allocations;

passes the offspring portfolio allocations through the dominance filter to identify a non-dominated subset of offspring portfolio allocations;

combines the non-dominated subset of parent portfolio allocations with the non-dominated subset of offspring portfolio allocations into a larger set of portfolio allocations;

passes the larger set of portfolio allocations through a non-crowding filter to identify a reduced subset of portfolio allocations;

creates a new population of individual portfolio allocations from the reduced subset of portfolio allocations;

updates the non-dominated portfolio allocations archive with the new population of individual portfolio allocations; and

passes the updated non-dominated portfolio allocations archive through the dominance filter to generate an interim efficient frontier in a portfolio performance space having at least three-dimensions, the interim efficient frontier being used in investment decisioning.

21. The computer readable medium of claim 20, wherein the non-dominated subset of parent portfolio allocations has a first cardinality.

22. The computer readable medium of claim 21, wherein the non-dominated subset of offspring portfolio allocations has a second cardinality that is different than the first cardinality.

23. The computer readable medium of claim 22, wherein the larger set of portfolio allocations has a third cardinality that is equal to the first cardinality plus the second cardinality.

24. The computer readable medium of claim 23, wherein the reduced subset of portfolio allocations has a fourth cardinality that is less than the third cardinality.

25. The computer readable medium of claim 20, wherein the reduced subset of portfolio allocations has a fourth cardinality, and wherein the new population of individual portfolio allocations is exactly equal to the reduced set of portfolio allocations if the fourth cardinality is equal to the initial population of individual portfolio allocations.

26. The computer readable medium of claim 20, wherein the reduced subset of portfolio allocations has a fourth cardinality, and wherein the new population of individual portfolio allocations is created by randomly drawing additional individual portfolio allocations from the portfolio allocations archive if the fourth cardinality is less than to the initial population of individual portfolio allocations.

27. The computer readable medium of claim 20, wherein the reduced subset of portfolio allocations has a fourth cardinality, and wherein the new population of individual portfolio allocations is created by randomly discarding individual portfolio allocations from the reduced subset of portfolio allocations if the fourth cardinality is greater than to the initial population of individual portfolio allocations.

28. The computer readable medium of claim 27, wherein the new population of individual portfolio allocations is created by randomly injecting individual portfolio allocations from the portfolio allocations archive until the fourth cardinality is equal to a desired number of individual portfolio allocations.