| 李 玮
杨红梅
王 浩
贾婧怡
刘 琪.核磁共振氢谱-PCA-SVM回归法用于稀奶油中植脂奶油掺假定量分析[J].中国油脂,2020,45(1):38~42.[LI Wei
YANG Hongmei WANG Hao
JIA Jingyi
LIU Qi.Determination of cream adulterated with non-dairy whip topping using1H NMR combined with principal component analysis and supportvector machines[J].China Oils and Fats,2020,45(1):38~42.] |
| 核磁共振氢谱-PCA-SVM回归法用于稀奶油中植脂奶油掺假定量分析 |
| Determination of cream adulterated with non-dairy whip topping using1H NMR combined with principal component analysis and supportvector machines |
| |
| DOI：10.12166/j.zgyz.1003-7969/2020.01.009 |
| 中文关键词: 核磁共振氢谱(1H NMR) 主成分分析 支持向量机 稀奶油 植脂奶油 掺假定量 |
| 英文关键词:1H nuclear magnetic resonance spectroscopy (1H NMR) principal component analysis (PCA) support vector machines (SVM) cream non-dairy whip topping quantification of adulteration |
| 基金项目:国家重点研发计划（2018YFC1602304） |
|
| 摘要点击次数: 1359 |
| 全文下载次数: 529 |
| 中文摘要: |
| 基于核磁共振氢谱-PCA-SVM回归方法，建立稀奶油中掺假植脂奶油的快速定量方法。核磁共振氢谱数据经分段积分、归一化等数据预处理后，利用PCA进行数据降维，采用交叉验证的方法对SVM中的参数进行优化，然后使用最优参数建立稀奶油中掺假植脂奶油比例的定量校准模型。将校正结果与PLS和SVM算法比较，并将所建立的3个模型用于测试集样本的预测。结果表明：基于PCA-SVM算法定量模型的RMSECV为3.69，RMSEP为5.87，训练集和测试集的R2分别为0.987 5和0.974 3，模型的稳定性、准确性以及模型的预测能力均优于PLS、SVM算法，且运行速度明显快于SVM算法。研究所建立的PCA-SVM模型表现出较好的模型稳定性和预测精度，结合核磁共振氢谱技术可以快速、准确地测定稀奶油中掺假植脂奶油含量，为规范市场上奶油蛋糕等烘焙制品的质量监管提供技术支持。 |
| 英文摘要: |
| The adulteration of non-dairy whip topping in cream was analysed using 1H NMR combined with principal component analysis (PCA) and support vector machines (SVM).1H-NMR was subsection integrated and normalized, and spectral dimension was also reduced through PCA. The cross validation was applied to optimize the parameters of PCA-SVM. The calibration model of determination of cream adulterated with non-dairy whip topping was established using the optimal parameters of PCA-SVM.The performance of PCA-SVM models was compared with partial least squares (PLS)and SVM, and the feasibility of these three methods was examined on the testing set. The results showed that the RMSECV and RMSEP obtained for PCA-SVM were 3.69 and 5.87 respectively, and the determination coefficients of the training set and testing set were 0.987 5 and 0.974 3 respectively. The stability, accuracy and prediction ability of the model were better than PLS and SVM algorithm, and the running speed was faster than SVM algorithm. In conclusion, a combination of 1H NMR with PCA-SVM method could quickly and accurately determined the content of adulterated non-dairy whip topping in cream, which could provide technical support for regulating the quality supervision of bakery products such as cream cake on the market. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
