Considering poor quality and high noise in medical data,we propose an ensemble classifier based on support vector machine (SVM) and apply this classifier to health identification.Health identification is one of the most important issues in public health studies.However,medical studies nowadays utilize basic classifiers to identify specific disease rather than health status.Samples in this study are composed by color features extracted from tongue image and health status label given by medical examinations.Theoretical analysis shows that ensemble approach has ability to improve performance of classifiers.Experimental results show that our classifier reduces test error rate with respect to several basic classifiers and existing ensemble classifiers.The reduction of test error indicates that the application of ensemble classifier on health identification is effective.Health identification provides a good foundation to further studies such as disease prevention.