改进支持向量机在睡眠分期预测模型中的应用

Application of Improved Support Vector Machine in Predicting Sleep Stages Classification

摘要: 睡眠分期是评估人类睡眠质量和诊断相关疾病的关键，对于睡眠分期预测的研究已有诸多成果。文中以脑电信号作为睡眠分期的工具，在支持向量机(Support Vector Machines, SVM)分类应用于睡眠分期预测模型的研究基础上，为了减少睡眠分期预测模型的建立的时间，采用K边界近邻法(K Nearest Bound Neighbor, KNBN)支持向量预选取的方法构造支持向量候选集，建立基于KNBN-SVM的睡眠分期预测模型。实验结果表明，该睡眠分期预测模型的预测准确度理想，并且耗时大幅度缩短。KNBN-SVM方法有效地改进了基于标准SVM睡眠分期预测模型，具有实用价值。

Abstract: Sleep staging is the key to assess human sleep quality and diagnose related diseases. There have been many results in the study of sleep state recognition. In this paper, EEG signals are used as a tool for sleep staging. Based on the research of support vector machine classification applied to sleep stages classification models, the K nearest boundary neighbor support vector pre-selection method is used to construct support vector candidate set in order to reduce the time to establish a model, build a sleep stage prediction model based on KNBN-SVM. Experimental results show that the prediction accuracy of the sleep staging prediction is ideal, and the time-consuming is greatly shortened. The KNBN-SVM effectively improves the sleep staging prediction based on the standard SVM, and has practical value.