摘要
为考察支持向量回归(SVR)在鱼糜近红外光谱(NIRS)分析中应用的可行性,采用SVR对73份带鱼糜样品的NIR漫反射光谱及其蛋白质和水分含量的化学测定数据进行处理,建立了蛋白质和水分含量NIRS定标模型,首次尝试将近红外光谱技术应用到带鱼糜主要成分含量的测定中。采用留一法交叉验证(LOOCV)的蛋白质和水分相关系数分别为0.90和0.96,并用独立样本集对模型进行外部验证。结果表明,SVR模型的预测能力比较好。因此,可以将NIR光谱支持向量回归法预测带鱼糜蛋白质和水分含量应用到鱼糜品质的快速评价中。
To observe the feasibility of the application of Support Vector Regression（SVR） in surimi by near infrared spectroscopy（NIRS） analysis,NIR diffuse reflectance spectra and experimental values of protein and moisture content of 73 surimi samples were processed with SVR,and NIRS calibration models of protein and moisture content were established.This is the first attempt to apply near infrared spectroscopy with determination of major components in Trichiurus japonicus surimi.The results of internal and external verification with leave-one-out（LOOCV） and independent samples showed that SVR model＇s predictive ability was good,the correlation coefficient of protein and moisture is 0.90 and 0.96 in LOOCV.Therefore,NIR spectroscopy with SVR can predict protein and moisture content in surimi and can be applied to the rapid evaluation of the quality of surimi.
出处
《计算机与应用化学》
CAS
CSCD
北大核心
2010年第12期1621-1624,共4页
Computers and Applied Chemistry
基金
"十一五"国家科技支撑计划项目(2008BAD94B09)
国家自然科学基金项目(30901125)
农业部"948"项目(2006-G43)
上海市教委重点学科建设项目(J50704)
关键词
带鱼糜
近红外光谱
支持向量回归
水分
蛋白质
Hairtail surimi
near-infrared spectroscopy
Support Vector Regression
moisture
protein
