An enhanced Cauchy mutation grasshopper optimization with trigonometric substitution: engineering design and feature selection

Selecting a subset of important features from a high-dimensional dataset is an important prerequisite for data mining. Meta-heuristic algorithms have gained attention in this field in recent years. The grasshopper optimization algorithm (GOA) is a meta-heuristic algorithm recently proposed based on the migration and hunting of grasshoppers in nature. However, the method suffers from a low diversity of the agents, which results in the stagnation problems, or immature convergence. To make GOA more competent in various situations, this paper stabilizes an improved GOA with new exploratory and exploitative features, which we have called it the SCGOA. The mechanism and structure of the proposed SCGOA are mainly divided into two steps: First, to balance the exploration and exploitation stages, trigonometric substitution is utilized for perturbation of the updating (evolution) of the position vectors of the individuals. Secondly, the diversity of the population is boosted using can Cauchy mutation-based strategy, which can help the grasshopper population to avoid the stagnation and lazy convergence. Therefore, Cauchy mutation is introduced to assist in an adequate variety of the position of the grasshopper population. Performance of SCGOA was validated on the latest IEEE CEC2017 benchmark functions in comparison with several well-known meta-heuristic algorithms. Various extensive results reveal that the proposed SCGOA has achieved a significant advantage over the other rivals. Finally, the Cauchy mutation-based SCGOA was also used for tackling four engineering design problems, and the results showed that SCGOA was superior to some state-of-the-art algorithms. We also developed the binary version of Cauchy mutation-based SCGOA in dealing with many feature selection datasets. The results on feature selection reveal that the binary version can outperform original GOA and other optimization algorithms, with higher classification accuracy, smaller error rate, and less number of features. We think the proposed optimizer can be widely tool for solving forms of the optimization problems. The research will be supported by open access materials and web service for any user guide at https://aliasghaheidari.com .