Noninvasive arteriosclerosis detection by sparse Least squares Support vector Machine

In this paper, we study the method of sparse least squares support vector machine(SL-SVM) to identify atherosclerosis. The time domain feature, frequency domain feature and wavelet feature of pulse wave are extracted as feature vectors. The dimension disaster problem is overcome by sparse feature vector method. VC dimension theory is used to control the VC dimension of data set by adjusting the parameters of kernel function, so as to predict the risk of model Identification method. Using 200 sets of pulse wave data in MIMIC-III Waveform Database Matched Subset establishes the diagnosis model of pulse wave and arteriosclerosis. The experimental results show that the model could realize the safe and non-invasive detection of atherosclerosis through self- learning, and the average recognition rate is 92.5%, which verifies the effectiveness of the method.