Research of Meridian Diagnosis System Based on SVM

Aiming at the deficiency of the current meridian diagnosis algorithms, SVM is applied to meridian diagnosis system. The system structure is described firstly, then the model selection of SVM is discussed in detail by taking chronic pharyngitis as an example: one-against-one method is used to realize multi-class; the problem of non-symmetrical samples of C-SVM is solved by giving positive and negative samples of different weights; a margin-based bound on generalization method is used to search parameters of the model. Finally, test results show that the classifier, which is realized and tested using vc++6.0, possess a very high recognition rate and can be applied to meridian diagnosis system.