A Modified Mean Gray Wolf Optimization Approach for Benchmark and Biomedical Problems

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Machine Learning

Background:

• Gray Wolf Optimization (GWO) is a nature-inspired meta-heuristic algorithm.
• Existing GWO variants may face challenges in complex optimization problems and classification tasks.
• Need for improved algorithms that balance exploration and exploitation for better solution finding.

Purpose of the Study:

• To develop a modified variant of the Gray Wolf Optimization algorithm, termed Mean Gray Wolf Optimization (MGWO).
• To evaluate the performance of MGWO on standard benchmark test functions.
• To assess the feasibility and effectiveness of MGWO in data classification tasks.

Main Methods:

• Modification of the position update (encircling behavior) equations of the standard GWO algorithm.
• Testing the proposed MGWO algorithm on 23 unimodal, multimodal, and fixed-dimension multimodal benchmark functions.
• Application of MGWO to the classification of 5 diverse datasets.

Main Results:

• MGWO demonstrated superior performance in solving benchmark test functions compared to GWO and Particle Swarm Optimization (PSO).
• The algorithm achieved high accuracy in the classification of 5 datasets.
• MGWO showed improved avoidance of local optima, leading to better solution quality.

Conclusions:

• The Mean Gray Wolf Optimization algorithm is an effective enhancement of the original GWO.
• MGWO offers a promising approach for both complex optimization problems and practical classification applications.
• The modified algorithm provides a robust alternative to other meta-heuristic methods, achieving better accuracy and local optima avoidance.