Machine Learning Meets Meta-Heuristics: Bald Eagle Search Optimization and Red Deer Optimization for Feature...

Area of Science:

• Bioinformatics
• Computational Biology
• Medical Informatics

Background:

• Microarray gene data presents high dimensionality challenges for accurate disease detection.
• Type II Diabetes Mellitus (DM) detection requires robust analytical methods for clinical utility.
• Existing classification methods may struggle with the complexity of genomic datasets.

Purpose of the Study:

• To evaluate feature extraction and selection methods for enhancing DM detection accuracy.
• To compare classifier performance with and without feature engineering on microarray data.
• To identify optimal techniques for improving diagnostic precision in Type II DM.

Main Methods:

• Feature extraction: Short-Time Fourier Transform (STFT), Ridge Regression (RR), Pearson's Correlation Coefficient (PCC).
• Feature selection: Meta-heuristic algorithms Bald Eagle Search Optimization (BESO) and Red Deer Optimization (RDO).
• Classification: Evaluated seven algorithms including Support Vector Machine with Radial Basis Function kernel (SVM-RBF) on microarray gene data.

Main Results:

• Pearson's Correlation Coefficient (PCC) with SVM-RBF achieved 92.85% accuracy without feature selection.
• BESO and PCC with SVM-RBF maintained high accuracy.
• Red Deer Optimization (RDO) combined with PCC and SVM-RBF yielded the highest accuracy of 97.14%.

Conclusions:

• Feature extraction and selection techniques substantially enhance Type II DM detection accuracy.
• Red Deer Optimization (RDO) coupled with Pearson's Correlation Coefficient (PCC) and SVM-RBF is highly effective.
• These optimized methods show significant potential for improving genomic-based diagnostics.