基于近红外光谱结合不同建模算法的防风色原酮含量预测定量模型研究  

RESEARCH ON THE QUANTITATIVE MODEL OF SAPOSHNIKOVIA DIVARICATA CHROMONE CONTENT PREDICTION BASED ON NEAR-INFRARED SPECTROSCOPY COMBINED WITH DIFFERENT MODELING ALGORITHMS

作　　者：焦慧平 裴媛媛 清明 布仁 莲花 额尔敦 白文明 

机构地区：[1]内蒙古医科大学药学院,内蒙古呼和浩特010059 [2]内蒙古医科大学蒙医药学院,内蒙古呼和浩特010059

出　　处：《内蒙古医科大学学报》2024年第6期561-567,共7页Journal of Inner Mongolia Medical University

基　　金：内蒙古自治区自然科学基金项目(2020MS08045);内蒙古自治区蒙医药协同创新中心科研项目(MYYXTPY202311);内蒙古医科大学面上项目(YKD2021MS022)。

摘　　要：目的 建立由防风近红外光谱预测色原酮含量的定量模型。方法 基于线性回归算法的主成分回归法(PCR)、偏最小二乘回归法(PLSR)和非线性回归算法的支持向量机回归法(SVMR),采用不同的光谱前处理方法优化模型的预测精度,建立定量模型并利用校正组样品和验证组样品对模型进行预测精度和模型稳定性的验证。结果基于线性回归算法建立的防风色原酮含量预测用PCR模型和PLSR模型的预测精度无显著差异,均能在较高的精度范围内对防风色原酮含量进行预测。基于非线性回归算法建立的SVMR模型预测精度较线性回归模型呈现更高的含量预测精度,最优模型的内、外部验证决定系数分别为0.971 6、0.989 5,校正均方误差和预测均方差分别为0.02%、0.01%,内、外部验证性能偏差比分别为5.41、6.76。结论 近红外光谱结合SVMR算法建立的定量预测模型的内、外部验证精度均较高且内、外部验证精度无显著差异,说明所建立定量模型的预测精度高、稳定性好。Objective To establish a quantitative model for predicting the content of Chromone by Saposhnikovia divari-cata near-infrared spectroscopy.Methods Based on linear regression algorithm of principal component regression(PCR),par-tial least squares regression(PLSR)and nonlinear regression algorithm of support vector machine regression(SVMR),different spectral pre-processing methods were used to optimise the prediction accuracy of the models,to build quantitative models and to validate the predictive accuracy and the model stability of the models using the calibration group samples and the validation group samples.Results There is no significant difference in the prediction accuracy between the PCR model and the PLSR model for predicting the Chromone content of Saposhnikovia divaricata based on the linear regression algorithm,and both of them can predict the Chromone content of Saposhnikovia divaricata with a higher accuracy range.The prediction accuracy of the SVMR model based on the nonlinear regression algorithm is higher than that of the linear regression model,and the internal and external verification determination coefficients of the optimal model are 0.9716 and 0.9895,respectively,the root mean squared error of calibration and the root mean squared error of prediction are 0.02%and 0.01%,and the internal and external relative prediction deviation are 5.41 and 6.76.Conclusions The internal and external verification accuracy of the quantitative prediction model established by NIR spectroscopy combined with SVMR algorithm is high,and there is no significant difference between the internal and external verification accuracy,indicating that the quantitative model has high prediction accuracy and good stability.

关 键 词：防风 色原酮含量 主成分回归法 偏最小二乘回归法 支持向量机回归法 

分 类 号：R917[医药卫生—药物分析学]