Portfolio optimization using genetic algorithm technique in utility maximization framework

Abstract

Markowitz famous formulation of reward-risk ratio to capture the most efficient portfolio allocation presupposes that risk is captured by the standard deviation of stock returns, the reward is captured by expected stock returns, and these two are the only considerations to be pondered about while deciding about optimal investment basket. If this were indeed so, we would have found all investors to hold to the same most efficient basket of the portfolio. Since findings, in reality, show that investors differ considerably in their portfolio allocation, it is easy to deduce that the criteria of risks and rewards are investor specific, and can not be generalized. In this context, it is assumed that each investor has his own set of criteria and utility curves attached to each of the criteria. The investor is set to maximize his global utility (which is a weighted average of individual utility values of the criterion dimensions) through an optimal portfolio allocation. The report is an attempt to optimize investor specific utility value using genetic algorithm search technique.