Improved Binary Grasshopper Optimization Algorithm for Feature Selection Problem

Area of Science:

• Computational Intelligence
• Metaheuristic Optimization
• Swarm Intelligence

Background:

• The grasshopper optimization algorithm (GOA) is a nature-inspired metaheuristic algorithm simulating grasshopper behavior.
• The binary grasshopper optimization algorithm (BGOA) adapts GOA for binary optimization problems.
• Existing BGOA methods may have limitations in exploration capability and solution quality.

Purpose of the Study:

• To enhance the exploration capability and solution quality of the binary grasshopper optimization algorithm (BGOA).
• To introduce novel modifications to the BGOA's step size and propose new transfer functions.
• To validate the effectiveness of the improved BGOA through comparative experiments and real-world applications.

Main Methods:

• Modification of the step size parameter within the BGOA framework.
• Development and integration of three new transfer functions to improve solution space exploration.
• Comparative analysis against BGOA, particle swarm optimization (PSO), and binary gray wolf optimizer (BGWO) on 23 benchmark functions.
• Application of the improved algorithm for feature selection on 23 UCI datasets.
• Statistical validation using Wilcoxon rank-sum and Friedman tests.

Main Results:

• The improved BGOA demonstrated significantly superior performance across most benchmark test functions compared to BGOA, PSO, and BGWO.
• The optimized algorithm achieved higher accuracy and selected fewer features in the feature selection task on UCI datasets.
• The proposed modifications effectively enhanced the algorithm's exploration capability and the quality of the obtained solutions.

Conclusions:

• The enhanced BGOA with modified step size and new transfer functions represents a significant improvement over existing binary optimization algorithms.
• The improved algorithm is highly effective for both general optimization problems and practical applications like feature selection.
• The study validates the proposed enhancements through rigorous testing and statistical analysis, confirming its superior performance and efficiency.