Optimizing feature selection and remote sensing classification with an enhanced machine learning method

In the era of rapidly expanding data volumes, the increasing dimensionality of features presents significant computational challenges, often degrading the performance of algorithms. Feature selection has emerged as a critical pre-processing step across various applications, aiming to identify and retain the most relevant features from datasets to enhance efficiency and accuracy. This study introduces an advanced wrapper-based feature selection approach, addressing key limitations of the original Grasshopper Optimization Algorithm (GOA), such as premature convergence and entrapment in local optima. The proposed Grasshopper Optimization Algorithm Harris Hawks Optimizer L & eacute;vy Flight (GHL) integrates two strategies: L & eacute;vy flight, which enhances the exploration phase by directing GOA toward promising regions of the search space, and Harris Hawks Optimizer techniques, which strengthen the exploitation phase to improve solution quality. Through three comprehensive experiments, the GHL algorithm demonstrated superior performance over nine comparative methods. The first experiment validated its efficacy in solving global optimization problems, achieving the best fitness values in most of the test functions. The second experiment highlighted its ability to effectively select relevant features across twenty benchmark datasets, achieving the best accuracy in 80% of the datasets. The third experiment applied GHL to remote sensing image classification, improving classification accuracy and yielding robust optimization outcomes.