变量筛选在茶叶咖啡碱近红外光谱定量分析模型中的应用  

Application of variable selection method in model of quantitative analysis for tea caffeine by near infrared

作　　者：邵美丽[1] 陈斌[1] 田萍[1] 

机构地区：[1]江苏大学食品与生物工程学院,镇江212013

出　　处：《安徽农业大学学报》2013年第2期262-265,共4页Journal of Anhui Agricultural University

基　　金：国家自然科学基金(31171697)资助

摘　　要：研究利用近红外光谱分析技术定量测定茶叶中咖啡碱的含量,目的是通过变量筛选简化模型并提高预测精度。试验中以135个来自大闽食品公司的茶叶作为研究对象,利用基于小波系数蒙特卡罗无信息变量消除法(WT-MC-UVE)进行变量筛选并结合偏最小二乘法(PLS)建立咖啡碱定量分析模型,选择交互验证均方根误差(RMSECV)和预测集均方根误差(RMSEP)以及预测相关系数(Rp)作为模型的评价指标。应用WT-MC-UVE筛选的90个变量所建立的模型,交互验证均方根误差,预测卷均方根误差,预测相关系数分别为0.124 8、0.1611和0.957 4。结果表明,该方法有效可行。In this research,we tested the content of tea caffeine by near-infrared（NIR） spectroscopy to simplify the model and increase the prediction accuracy by a method of variable selection.One hundred and thirty-five tea samples from Damin Food Company were tested.Monte Carlo uninformative variables elimination based on wavelet coefficient（WT-MC-UVE） method was used for variable selection,and the model of quantitative analysis for tea caffeine was established by partial least squares（PLS）.The root mean square error of cross validation（RMSECV）,the root mean square error of prediction se（t RMSEP）and correlation coefficients（Rp） were chosen for the appraisal criterion of the model.Ninety variables were optimized for modeling,and the model’s RMSEC,RMSEP and Rp were 0.1248,0.1611 and 0.9564,respectively.The results show that this method is valid and feasible.

关 键 词：近红外 咖啡碱 WT-MC-UVE 变量筛选 

分 类 号：TS272[农业科学—茶叶生产加工] O657.33[轻工技术与工程—农产品加工及贮藏工程]