HGAB3C: A new hybrid global optimization algorithm

Abstract

This paper proposes a new optimization algorithm, namely HGAB3C, and presents its performance on the CEC-2014 test suite. In HGAB3C, simple genetic algorithms (GAs) and big bang-big crunch (BB-BC) are hybridized. The algorithm carries out global searches using a simple GA. In every generation the BB-BC algorithm is used to carry out local searches. The addition of local search has improved the capability of simple GAs significantly. The performance of the proposed algorithm is compared with 17 other optimization algorithms on all 30 functions of the CEC-2014 benchmark suite. It is observed that HGAB3C outperforms all other algorithms on 4 benchmark functions. For the 3 other functions, its performance equaled the best of the competing algorithms, which makes HGAB3C’s performance best in a total of 7 benchmark functions. Out of the 18 competing algorithms, the proposed algorithm ranked second for the unmatched best mean error measure. For the best performance measure (number of functions giving unmatched best and equaled best mean error), the proposed algorithm was the third best. As far as the speed of convergence is concerned, the algorithm gave an unmatched best performance for the shifted Schwefel function (function 10 of CEC-2014 test bench). It obtained a mean error value of 0.00E+00, outperforming the previous best of 1.23E-03, converging to the target result in an average of 346.44 generations, which no other algorithm could achieve.