基于近红外光谱的杏仁糖中糖含量预测模型的研究

STUDY ON PREDICTION MODEL OF SUGAR CONTENT IN MARZIPAN BASED ON NEAR INFRARED SPECTROSCOPY

本研究以32个杏仁糖样品为研究对象,使用4种不同近红外光谱仪采集样品的光谱数据,利用二阶导数光谱法处理数据,并结合Savizky-Golay平滑方法,计算二阶导数光谱。将数据集预处理后,建立偏最小二乘(PLS)和支持向量机(SVM)模型用于预测未知样品的糖含量,通过对两个模型的应用发现,PLS模型预测值与真实值偏差较大,均方根误差为2.9822,而SVM模型中利用10折交叉验证优化参数,优化参数后预测值几乎全部与真实值相同,预测值与真实值间均方根误差为0.0127,误差极小。综上所述,SVM模型均方根误差较小,所以选择SVM模型作为糖的预测模型,为杏仁糖样品中糖含量的快速检测提供一种精确简单的方法,此模型可推广至食品中糖含量的定量分析。

In this study, 32 marzipan samples were taken as the research object. Four kinds of near-infrared spectrometers were used to collect the spectral data of the samples. The second derivative spectrum was processed by the second derivative spectrum method, and the second derivative spectrum was calculated by the savizky-Golay smoothing method. After preprocessing the data set, partial least squares(PLS) and support vector machine(SVM) models are established to predict the sugar content of unknown samples. Through the application of the two models, it is found that the deviation between the predicted value and the real value of PLS model is large, and the root mean square error is 2.9822. In SVM model, 10 fold cross validation is used to verify the optimization parameters. After the optimization parameters, the predicted value is almost the same as the real value, and the root mean square error between the predicted value and the real value is 0.0127, with minimal error. Therefore, SVM model is selected as the prediction model of sugar, which provides a simple and accurate method for the rapid detection of sugar content in almond marzipan samples. This model can promote the quantitative analysis of sugar content in food.