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

Abstract

BACKGROUND: Heart rate variability (HRV) can reflect the relationship between heart rhythm and sleep structure. OBJECTIVE: In order to study the effect of support vector machine (SVM) on the results of automatic sleep staging and improve the effectiveness of heart rate variability (HRV) as a sleep structure biomarker, thereby realize long term and non-contact monitoring of sleep quality. METHODS: Two kinds of parameter optimization methods are applied to stage sleep experiments when the known SVM can be used for automatic sleep staging. By factor analysis of the time domain, frequency domain, and nonlinear dynamic characteristics of subjects' HRV signals, the accuracy of the cross-validation method (K-CV) is used as the fitness function value in genetic algorithm (GA) and particle swarm optimization (PSO). Furthermore, GA and PSO are used to optimize the SVM parameters. RESULTS: The results show that the accuracy rate of sleep stage is 64.44% when parameters are not optimized, the accuracy rate based on PSO is improved to 78.89% and the accuracy rate based on GA is improved to 84.44%. CONCLUSION: Both optimization algorithms can improve the accuracy of SVM for sleep staging and better results based on GA in the experiment.