Determination of the Control Parameters of Evolutionary Techniques Utilizing Analysis of Variance

Evolutionary techniques are considered to be an alternative to the effective solution of many, especially non-standard optimization problems. However, these algorithms are controlled by a special set of control parameters according to algorithm type and these settings radically affect the quality of the results. Paper focuses on the possibility of using selected statistical methods to determine the effective values of control parameters, while algorithm of differential evolution has been chosen as an evolutionary technique representative. The analysis has been elucidated on an illustrative example of Sortino ratio maximization problem which is one of the nonlinear measures of portfolio performance.