Asarum subspecies identification with pattern recognition techniques

Chinese medicine pharmacologists examine the chemical features of Chinese medicine materials for identification of the subspecies. In this paper, three different types of chemical data, namely main constituent content, inorganic element content and HPLC fingerprint data of 54 asarum samples are tested and analyzed. Some types of data with strong connection with the sample subspecies classification are firstly filtered out with Principle Component Analysis and separability measure. Chemical features of these data types are then ranked with concern of the correlation with the sample subspecies using SVM RFE. At last, the effect of the filtered out chemical features on the sample subspecies classification are verified using leave-one-out strategy.