Parameter tuning of GSA using DOE

Parameter tuning has critical influences on the performance of evolutionary algorithms. Deliberate parameter investigation and changing the value of them is very expensive and time consuming. This paper has applied Design of Experiment (DOE) method to tune the parameters of Gravitational Search Algorithms (GSA) systematically. Also, to reduce its complexity and increase the performance, simple modification has been presented to determine the number of effective objects (Kbest). Best configurations of 17 standard functions are obtained by executing DOE. Analysis of the results confirms that parameter tuning and Kbest modification have improved the performance of the GSA. Meanwhile, these results have been obtained by least experiments in acceptable time.