| Abstract |
In this paper, we proposed Joint Opposite Selection (JOS) operator that is a joint of two opposition learning techniques: the Selective Leading Opposition (SLO) and the Dynamic Opposite (DO). SLO uses a linearly decreasing threshold value to determine the close distance dimension of the search agents. DO provides the search agents chances to expand their abilities in the search space. We applied JOS to the Harris Hawks Optimization (HHO), the performance is increased because JOS balances the capability of exploration phase by using SLO and exploitation phase by using DO. The new algorithm, named Harris’ Hawks Optimization-Joint Opposite Selection (HHO-JOS), is also proposed in this research as an enhanced version of HHO to solve single-objective problems. When the hawks deploy JOS, SLO assists the hawks to succeed in exploitation phase by changing their close distance dimension and DO tries to diverse the search space range of the hawks in the exploration phase using a Random Jump Strategy (RJS). The sufficient Jumping rate of DO in HHO-JOS is 0.25, according to our experimental results. The proposed algorithm was included in a competition conducted on 30 benchmark functions of CEC 2014 and 29 benchmark functions of CEC 2017. Both benchmarks contain collections of single-objective problems for real parameter numerical optimization. The problems were employed to evaluate and compare the proposed HHO-JOS to the original HHO, three variations of OBLs embedded in the original HHO, and 31 nature-inspired algorithms by using a scoring metric. The results of the competition showed that the premiere JOS on HHO consistently achieves robustness performance on CEC 2014 and CEC 2017. Comprehensive statistical analysis also demonstrated that HHO-JOS can compete with many leading optimization algorithms. Therefore, we can conclude that the proposed joint opposite selection is well-matched to HHO and succeeds in elevating HHO-JOS. |
Publication
Journal Publication
ไทย