Abstract
A variety of bond portfolio optimization problems of institutional investors are formulated as linear and/or bilinear fractional programming problems and algorithms to solve this class of problems are discussed. Our objective is to optimize certain index of returns subject to constraints on such factors as the amount of cash flow, average maturity and average risk, etc. The resulting objective functions and constraints are either linear, bilinear or bilinear fractional functions. T. he authors devised a special purpose algorithm for obtaining a local optimal solution of this nonconvex optimization problem containing more than 200 variables. Though it need not generate a global optimum, it is efficient enough to meet users' requirement.
