An accelerated sine mapping whale optimizer for feature selection.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Machine Learning

Background:

• Global optimization problems require efficient algorithms to explore vast solution spaces.
• The standard Whale Optimization Algorithm (WOA) faces challenges in balancing exploration and exploitation.
• Enhancing WOA's performance is crucial for complex optimization and feature selection tasks.

Purpose of the Study:

• To introduce an improved Whale Optimization Algorithm (SWEWOA) for global optimization.
• To develop a novel feature selection method (BSWEWOA-KELM) using the enhanced algorithm.
• To evaluate the effectiveness of SWEWOA and BSWEWOA-KELM on benchmark datasets.

Main Methods:

• Population initialization using Sine mapping strategy (SS).
• Incorporation of Escape Energy (EE) to balance exploration and exploitation.
• Integration of Wormhole Search (WS) to improve exploitation capabilities.
• Development of a binary SWEWOA combined with Kernel Extreme Learning Machine (KELM) for feature selection.

Main Results:

• SWEWOA demonstrated superior performance against 26 comparison algorithms on CEC 2017 and 2022 test sets.
• BSWEWOA-KELM outperformed 8 high-performance algorithms in feature selection across 16 diverse datasets.
• The hybrid approach significantly enhanced optimization and feature selection capabilities.

Conclusions:

• The proposed SWEWOA effectively addresses global optimization challenges.
• BSWEWOA-KELM provides an excellent feature selection method for classification problems.
• The hybrid strategies significantly improve the performance of the Whale Optimization Algorithm.