Identification of astragalus origin based on terahertz spectroscopy and machine learning

Astragalus is commonly used Chinese medicine materials, which has the effect of invigorating qi. Due to the different origin, the quality and efficacy of astragalus are different, which has an important impact on clinical application. This paper presented a new identification method of astragalus origin based on terahertz time-domain spectroscopy (THz-TDS) combined with machine learning. Five astragalus from Xinjiang, Shanxi, Inner Mongolia, Gansu and Qinghai were taken as the research objects. The terahertz spectral characteristics of astragalus from different producing areas were described, and the discriminant classification models were established by combining the absorption coefficient spectra in the range of 0.2~1.3THz with random forest (RF), backpropagation neural network (BPNN) and convolutional neural network (CNN). The results showed that CNN had the best prediction effect on astragalus origin, and the overall accuracy reached 95.2381%, indicating that terahertz time-domain spectroscopy combined with convolutional neural network model could be effectively used for automatic identification of astragalus origin, providing a new method reference for identification of Chinese herbal medicine origin.