Comparison of support vector machines based on particle swarm optimization and genetic algorithm in sleep staging

Area of Science:

• Biomedical Engineering
• Computational Neuroscience
• Sleep Medicine

Background:

• Heart rate variability (HRV) is a physiological indicator reflecting the interplay between cardiac rhythm and sleep architecture.
• Accurate sleep staging is crucial for diagnosing sleep disorders and assessing overall health.
• Non-contact, long-term sleep monitoring methods are highly desirable for clinical and research applications.

Purpose of the Study:

• To evaluate the impact of support vector machine (SVM) parameter optimization on automatic sleep staging accuracy.
• To enhance the utility of heart rate variability (HRV) as a biomarker for sleep structure.
• To enable long-term, non-contact sleep quality assessment.

Main Methods:

• Employing genetic algorithm (GA) and particle swarm optimization (PSO) for SVM parameter tuning in sleep staging.
• Utilizing factor analysis on time, frequency, and nonlinear HRV signal characteristics.
• Applying K-fold cross-validation (K-CV) accuracy as the fitness function for optimization algorithms.

Main Results:

• Unoptimized SVM achieved a sleep stage accuracy of 64.44%.
• PSO-optimized SVM improved accuracy to 78.89%.
• GA-optimized SVM demonstrated the highest accuracy at 84.44%.

Conclusions:

• Both GA and PSO effectively enhance SVM performance for automatic sleep staging.
• Genetic algorithm optimization provided superior results compared to particle swarm optimization in this study.
• Optimized SVM with HRV analysis offers a promising approach for accurate, non-contact sleep monitoring.