Comparative Study of Several Pattern Recognition Methods in the Identification of Volatile Oils of Tradition-al Chinese Medicine by Infrared Spectroscopy / 中国药房

ABSTRACT

OBJECTIVE:To compare the performance of several pattern recognition methods in the identification of volatile oils of traditional Chinese medicine(TCM)by infrared spectroscopy. METHODS:The volatile oils of several Lonicera and Citrus TCM were determined by infrared spectroscopy. All samples of infrared spectrum were classified by hierarchical clustering,K-mean clustering,artificial neural networks,and support vector machine. RESULTS:The results of hierarchical clustering and K-mean clus-tering were ineffective. Methods of artificial neural networks and support vector machine achieved correct classification rate of 100%. CONCLUSIONS:Artificial neural networks and support vector machine can be combined with infrared spectroscopy to cre-ate chemometric fingerprinting for the identification of volatile oils of TCM.