Enhancing the Harmony Search Algorithm Performance on Constrained Numerical Optimization

In this paper, an improved harmony search (ImHS) algorithm is presented. HS is a simple but efficient metaheuristic method explored in recent literature, that simulates the process of musical improvisation. Two modifications for parameter tuning are proposed to enhance the algorithm performance in the solution of constrained numerical optimization problems, maintaining the simplicity of its original design. Metaheuristics are methods for solving optimization problems, and are based in two processes: exploration (diversification) and exploitation (intensification). The proposed modifications improve both processes in HS, without breaking their balance. A well-known ideal problem set was used as a reference to compare the efficiency of the developed algorithm ImHS with HS and three of its most successful variants, and also with two other metaheuristics of different nature, artificial bee colony (ABC) and modified ABC (MABC). Various techniques were applied to evaluate the algorithm performance with the proposed modifications, in order to validate the reliability of the comparison. In most case studies, ImHS far surpassed the results of HS and ABC, also improving the performance of the selected variants. Additionally, its results reached a similar quality than the obtained with MABC but with a significantly lower computational cost, suggesting that it can be a useful tool for solving real-world optimization problems if they are modeled as constrained numerical cases.