Dilated Cardiomyopathy Metabolomics Data Classification Based on DAE-SVM Algorithm

A novel deep auto-encoder network combined with support vector machine (DAE-SVM) are prepared for the arrangement of metabolomics data. Because of their small sample size, high-dimensional, nonlinear and noisy parameters, traditional feature abstraction and classifications are very difficult to achieve satisfactory results. DAE performs non-linear transformations with hidden layers, which can learn complex relationships. It has a solid capability to show high-order appearance and can excerpt more complicated characteristics of metabolomic data. This manuscript considers Boltzmann Machine to complete the pre-training of DAE, the conjugate gradient was adopted to complete the fine-tuning, and SVM completes the classification. Empirical results on actual metabolomics data of expound cardiomyopathy concluded the proposed model has attained best accomplishment compared to other extant algorithms.