论文部分内容阅读
采用傅里叶变换红外光谱,测定了45个来自青海省不同产地的枸杞样品的红外光谱。小波变换对红外光谱原始数据进行了预处理。红外光谱数据压缩到原来的1/8,其分析精度与原始光谱数据基本相当。将45个样本数据分为30个训练集和15个测试集,建立随机森林(RF)预测枸杞产地模型,使用内部交叉验证和外部数据进行验证。采用R语言实现随机森林算法,并对模型的参数进行了优化。结果,所建立的判别模型中训练样本判别正确率为100%,测试样本判别正确率为100%。研究结果表明,建立的模型能够正确地对枸杞样品快速地进行产地鉴别,红外光谱法结合随机森林可作为中药材产域分类鉴别的一种新的现代化方法。
Fourier transform infrared spectroscopy was used to determine the infrared spectra of 45 samples of Lycium barbarum from different areas in Qinghai province. The original data of the infrared spectrum was preprocessed by wavelet transform. Infrared spectral data is compressed to the original 1/8, the analysis accuracy is basically the same as the original spectral data. The 45 sample data were divided into 30 training sets and 15 test sets to establish a random forest (RF) prediction model of wolfberry producing area, using internal cross-validation and external data validation. The random forest algorithm is implemented by R language, and the parameters of the model are optimized. The results show that the discriminant accuracy of the training samples in the discriminant model is 100% and that of the test samples is 100%. The results show that the established model can accurately identify the origin of Lycium barbarum, and infrared spectroscopy combined with random forest can be used as a new modern method of identification of Chinese herbal medicines.