Discussion on Statistical Pattern Recognition Model Related to Herbal Property and Lipid Components of Chinese Materia Medica / 世界科学技术-中医药现代化

ABSTRACT

This study was aimed to explore recognition models and to establish statistical pattern recognition methods of cold-hot property markers based on lipid components GC-MS chromatogram of Chinese materia medica (CMM). GC-MS fingerprints of lipid components contained in 60 kinds of cold or hot property of CMM were used as the research object. The database was established. Five types of model establishment strategies were compared. Optimal modeling patterns were screened for the identification of herbal property markers of lipid components GC-MS chromatogram. The results showed that support vector machine (SVM) was the best model to discriminate cold or hot property among 60 types of CMM, which were able to effectively mark the characteristic area. The strongest markers tending to cold property was at the retention time of 61.550 min, while the strongest markers tending to hot property was at the retention time of 31.395 min. It was concluded that cold or hot property of CMM had close relationship with lipid components. The lipid component was one of the material bases of CMM. The mathematical statistical model based on material base and herbal property can be used to identify and predict the cold and hot property of CMM.