主成分分析-近红外光谱法快速测定婴儿配方乳粉中必需脂肪酸

Study on the determination of essential fatty acids in infant milk powder based on principal component analysis of near infrared spectroscopy

利用主成分分析法将婴儿乳粉的近红外光谱数据与必需脂肪酸含量建立校正模型,并采用交互验证和外部检验两种方式来考察模型的可靠性。得到模型校正相关系数(R2)0.8704,均方估计残差(RMSEC)和交互校验均方残差(RMSECV)分别为0.6275和0.8907,应用所建立的必需脂肪酸近红外模型对婴儿乳粉必需脂肪酸含量进行预测,并将预测值和化学测量值进行配对t检验,结果表明两者无显著差异。

Principal component analysis prediction model for essential fatty acid content of the infant milk powder by near infrared spectroscopy had been established. Reliability of the model was verified by cross-validation and external-validation. Predictive correlation coefficient of the model for essential fatty acid content is 0. 8704, and root mean square error of calibration （RMSEC） is 0. 6275, while root mean square error of cross validation （RMSECV） is 0. 8907. By paired sample test, the results of prediction had no significant difference compared with those by GC.