A classification algorithm for TCM syndromes based on P-SVM

Based on traditional SVM, prior knowledge support vector machine (P-SVM) introduces application-oriented metrics into the training set to express expert knowledge. Developing with SLT theory, it is a new classification and prediction method established on firm mathematical foundation. Also, SVM provides the best solution of classification and prediction of limited sample set. In this paper, we introduce prior knowledge based P-SVM model into the software-developing project: Information Management System of TCM Syndrome, funded by the Guangdong Bureau of Traditional Chinese Medicine (TCM) Administration. After forming the rules from expert knowledge, we at first calculate the confidence values of each sample, and then use the sample set to train P-SVM by using P-SMO algorithm, which is a prior knowledge based improved version out of the traditional ones. Experiments show that our algorithm is effective. And the knowledge derived from TCM syndrome also confirms great accuracy of the classification process.