Symbiotic organisms search algorithm using random walk and adaptive Cauchy mutation on the feature selection of sleep staging

Sleep staging can objectively evaluate sleep quality to effectively assist in preventing and diagnosing sleep disorder. Because of the multi-channel and multi-model characteristics of physiological signals, high-dimensional features cannot be avoided when studying sleep staging. High-dimensional features are often mixed with redundant and irrelevant features, which may decrease the accuracy of classifiers and increase the computational cost. Feature selection can remove redundant and irrelevant features but is considered a challenging task in machine learning. Therefore, feature selection can be regarded as a multi-objective optimization problem. In this paper, the proposed symbiotic search algorithm (RCSOS), which is based on random walk and adaptive Cauchy mutation, can improve the optimization performance of the original algorithm. A binary version of RCSOS is proposed according to the twenty transformation functions. Then, the proposed algorithm is applied to feature selection in sleep staging. To validate the performance and generalization of the algorithm, seven groups of data from two different datasets were tested. Compared with the state-of-art algorithms, the proposed binary version of the RCSOS algorithm performs best on feature selection of sleep staging.

Automated summary

Sleep staging is crucial for preventing and diagnosing sleep disorders. However, the high dimensionality and abundance of redundant and irrelevant features in physiological signals pose challenges in studying sleep staging. This paper proposes an improved algorithm that enhances feature selection performance.