基于介电频谱的灵武长枣品质参数的预测模型

Prediction Models for Quality Parameters of ‘Lingwuchangzao' Jujube Based on Dielectric Spectra

为寻找预测灵武长枣品质的最优模型,以长枣的介电损耗因子?"和介电常数?’频谱进行内部品质参数(可溶性固形物、可滴定酸含量和含水率)的建模研究。通过遗传算法(genetic algorithm,GA)和相关系数(correlation coefficients,CC)法提取了介电谱的有效信息;采用偏最小二乘(partial least squares,PLS)、主成分回归(principal components regression,PCR)和支持向量机(support vector machine,SVM)法建立了品质参数的预测模型;以决定系数(R^2)、校正标准偏差和预测标准偏差等模型评价方法确定了品质参数的最优预测模型。结果表明:基于介电损耗因子?"建立的可溶性固形物含量、可滴定酸含量和含水率的最佳预测模型分别为GA-PCR、GA-PLS和GA-PLS,且R^2均达到0.9以上;基于介电常数?’建立的可溶性固形物含量、可滴定酸含量和含水率的最佳预测模型分别为CC-PLS、GA-SVM和GA-PLS,R^2达到0.8以上,且验证效果较好。本研究为利用介电频谱快速预测长枣品质提供了可靠的方法。

This study aimed to establish different models to predict some quality parameters(soluble solids,titratable acidand moisture contents)of‘Lingwuchangzao’jujube based on dielectric loss factor?’’and dielectric constant?’spectraand to select the optimal one.The effective information from dielectric spectra was extracted by genetic algorithm(GA)and the correlation coefficient method.The prediction models for quality parameters were established using partial leastsquares(PLS),principal components regression(PCR)and support vector machine(SVM).The optimal prediction modelwas determined by the coefficient of determination(R2),root mean square error of calibration set(RMESC)and root meansquare error of predication set(RMSEP).The results showed that the optimal prediction models for soluble solids,titratableacid and moisture contents were GA-PCR,GA-PLS and GA-PLS based on dielectric loss factor?’’spectra with a R2valueof greater than0.9,respectively.The best prediction models based on dielectric constant?’spectra were CC-PLS,GA-SVMand GA-PLS with a R2value of greater than0.8,respectively.These models were validated with satisfactory results.Inconclusion,this study can provide a reliable method for rapid prediction of jujube quality using dielectric spectra.