This study is to find the optimum classifier that can be easy and robust diagnostic method of the obstructive sleep apnea (OSA) using a heartbeat signal. The heartbeat signal was acquired from the 92 patients with OSA. The dataset consists 98,060 epochs, from them the training sets contained 68,642 epochs from the 63 OSA patients and test sets contained 29,418 epochs from the 29 OSA patients, respectively. The heartbeat signal was analyzed in the time and frequency domain and six features were extracted (normal-to-normal [NN], standard deviation of mean NN [SDNN], root mean square of successive differences [rMSSD], low-frequency [LF], high-frequency [HF], and LF/HF ratio). All extracted features were used to train the following classifiers: linear discriminant analysis (LDA), decision tree (DT), logistic regression (LR), k-nearest neighbor (KNN) and support vector machine (SVM). The top three classifiers (SVM, DT, and LDA) showed the accuracy of 93.2%, 93.2%, and 93.2% for test sets, respectively. Then, the top three classifiers could be effective on sleep studies and OSA detections.

In this paper, we compared the performance of support vector machine (SVM) and fuzzy SVM (FSVM) for reduction of learning time when classifying large-scale time series data into two classes. The fast learning time of the pattern classifier for large time series data is very useful in decision support systems. Considering the high interest in healthcare, including big data analysis, it is necessary to design a pattern classifier with a fast learning capability. We used large-scale time series data of 32 patients with sleep apnea (SA) for this study. The experiment was conducted by extending the parameter n, of the fuzzy membership function of FSVM, from 1 to 500. The result shows that the shortest learning time of FSVM is 3 s for radial base function (RBF), 17 s for a polynomial, and 35 s for a linear kernel, where the parameter n of the fuzzy membership function is n = 2, n = 433, and n = 4, respectively. The maximum classification hit rate of FSVM is 93.23%, and the learning time is significantly faster than conventional SVM. Therefore, FSVM can be used as a good classifier for the large-scale time series SA database.