Using Genetic Algorithms for a Portfolio Selection regarding to Nonlinear Goals (Tehran Stock Exchange)

Abstract

Generally, investors consider simultaneously conflicting objectives such as rate of return, risk and liquidity in portfolio selection. On the other hand, every investor has his own specific preferences about objectives. Therefore, we can use Multi Objective Decion Making (MODM) techniques in order to solve portfolio selection problem. The Studies shows that a MODM technique by nonlinear goals such as minimization of nonsystematic risk and skewness maximization isn’t employed for portfolio selection, so a new approach is applied. We employ MODM model to select a best portfolio in 50 superior companies in Tehran stock exchange with regards to optimization objectives of return, systematic risk, nonsystematic risk, skewness, liquidity and sharp ratio. This model is non-convexed, so operational research algorithms can not find the best solution; therefore we use Genetic Algorithms (GA) for achieving nonlinear multi-objective model. In the end, the result of GA is comprised with Markwitz classic model and goal programming (containing linear and nonlinear objectives). The comparison indicates that although return of the portfolio of GA model is less than the other models, but GA has the best results in decreasing risk criteria which completely cover the return and lead to best results. The other advantage of using GA is a higher diversification in its proposed portfolio in comparison with other models.

Keywords

• genetic algorithms
• nonlinear objective
• return
• risk
• portfolio selection