Exploring a financial product model with a two-population genetic algorithm

This paper describes a successful application of evolutionary computation to a difficult and commercially significant constrained optimization problem. The Financial Product model, for optimizing mortgage refinancing packages, is introduced. It is a realistic, and very challenging, optimization problem, for which standard solvers leave much to be desired. An untuned two-population genetic algorithm (GA) has been remarkably successful in finding good, feasible and nearly optimal solutions. In addition, the genetic solver provides important information for management decision making besides simply a good solution to the model. Finally, the paper undertakes a case study in order to investigate the details of how and why the two-population GA works.