An Enhanced Educational Competition Optimizer Integrating Multiple Mechanisms for Global Optimization Problems.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Machine Learning

Background:

• The original Educational Competition Optimizer (ECO) framework faces challenges with limited information exchange, slow convergence, and an unstable exploration-exploitation balance.
• These limitations hinder its effectiveness in solving complex optimization problems.

Purpose of the Study:

• To introduce an enhanced version of the ECO algorithm, termed EECO, designed to overcome the limitations of the original framework.
• To improve solution accuracy, convergence speed, and the stability of the exploration-exploitation ratio in optimization tasks.

Main Methods:

• EECO incorporates three key mechanisms: a regenerative population strategy using elite solution covariance for diversity, a Powell mechanism for accelerated exploitation, and a trend-driven update for adaptive exploration-exploitation balance.
• The algorithm was rigorously tested on 29 CEC-2017 benchmark functions and nine real-world constrained engineering problems.

Main Results:

• EECO significantly outperformed eight recent algorithms, including EDECO and LSHADE-SPACMA, in terms of solution accuracy and reduced standard deviations.
• The algorithm achieved consistently high rankings across various dimensionalities (10-D to 100-D) on CEC-2017 benchmarks and real-world engineering problems, demonstrating superior and scalable performance.
• Statistical significance of improvements was confirmed using the Wilcoxon rank sum test.

Conclusions:

• EECO presents a robust and effective enhancement to the ECO algorithm, offering remarkable convergence accuracy and reliable stability.
• Its dimension-scalable performance and superior results position EECO as a highly promising variant for advanced optimization challenges.