关键词: 红外光谱, 机器学习, 聚类分析, 神经网络

Abstract: There are great differences in the characteristics of near-infrared and mid-infrared spectra of different kinds of traditional Chinese medicine. Due to the different chemical components such as inorganic elements and organic substances, even if the origin of the same traditional Chinese medicine is different, the labeling effect under near-infrared and mid-infrared spectral irradiation will have different spectral characteristics which can be used to classify Chinese herbal medicine and identify the origin of Chinese herbal medicine. With the help of MATLAB software tool and K-means clustering algorithm in SPSS classification tool, unsupervised machine learning is carried out on traditional Chinese medicine to classify traditional Chinese medicine; Using SPSS neural network multilayer perceptron and the random forest algorithm provided by Python language, 70% of the data set is used as the training set and 30% as the verification set to train the supervised machine learning model which is finally used to identify and predict the origin of traditional Chinese medicine.

Key words: infrared spectrum, machine learning, cluster analysis, neural network