Ensemble mutation and multi-population-driven differential evolution for numerical optimization and feature selection...

Area of Science:

• Computational Biology
• Machine Learning
• Bioinformatics

Background:

• Breast cancer diagnosis is challenged by high-dimensional datasets, reducing predictive model efficiency.
• Effective feature selection is crucial for improving diagnostic accuracy and model performance.

Purpose of the Study:

• To develop an advanced differential evolution algorithm (EMMDE) for optimal feature selection in breast cancer diagnosis.
• To enhance predictive model efficiency and accuracy by addressing the challenges of high-dimensional medical data.

Main Methods:

• Developed an advanced differential evolution algorithm named ensemble mutation and multi-population-driven differential evolution (EMMDE).
• EMMDE incorporates population division, ensemble mutation rules, and a time-varied geometrically diversified scheme.
• Utilized Gaussian binning for a guiding vector to balance exploration and exploitation, promoting convergence and population diversity.
• Developed a binary version of EMMDE using a transfer function for validation on benchmark and breast cancer datasets.

Main Results:

• EMMDE demonstrated superior search ability on IEEE CEC2017 benchmark functions.
• The binary EMMDE algorithm achieved highly accurate and promising results for feature selection on UCI and breast cancer datasets.
• Experimental verification using MCE and other metrics confirmed the algorithm's effectiveness.

Conclusions:

• The proposed EMMDE algorithm significantly improves feature selection for breast cancer diagnosis.
• EMMDE offers a robust solution for handling high-dimensional datasets and enhancing predictive model performance.
• The study highlights the impact of novel strategies in metaheuristics for optimization and feature selection problems.