K-mean and Double Cross-Validation Algorithm for LS-SVM in Sasang Typology Classification

The Sasang typology is the traditional typology theory in Oriental Medicine. The medical typology distributes people into four types based on their traits: Tae-Yang, So-Yang, Tae-Eum and So-Eum type. In the paper, we design a model for Sasang typology classification based on the least squares support vector machines (LS-SVM). We use k-mean algorithm for decision the feature index from the side face, then we add the side face ratios into our feature space for improvement. The choice of resample method is very important for parameters' optimization in SVM and it will influence the system's stability. We propose a novel resampling algorithm called double cross-validation through comparing the cross-validation with bootstrap approach. The result shows that the established model under double cross- validation algorithm based on LS-SVM is more robust with high performance and suffices for the requirements of control and optimization for classification processes.