This paper proposes a new model for portfolio selection in which the expect
ed returns of securities are considered as variables rather than as the ari
thmetic means of securities. A genetic algorithm is designed to solve the o
ptimization problem which is difficult to solve with the existing tradition
al algorithms due to its nonconcavity and special structure. We illustrate
the new model by a numerical example and compare the results with those der
ived from the traditional model of Markowitz.
Scope and purpose
Portfolio selection, originally articulated by Markowitz, has been one of t
he most important research fields in modern finance. Several new models and
extensions such as the inclusion of transaction costs and taxes have been
proposed to improve the performance of portfolio investment. All those mode
ls and extensions have advantages and disadvantages in both theory and prac
tical applications. The purpose of this paper is to describe the return and
risk of a portfolio more accurately. On the basis of an order of expected
returns of securities, we propose a new model for portfolio selection. (C)
2000 Elsevier Science Ltd. All rights reserved.