基于PCA-SVMR快速测定复方氯丙那林和对乙酰氨基酚

Fast determination of compound clorprenaline and compound paracetamol based on PCA-SVMR

基于主成分分析-支持向量机回归(PCA-SVMR)方法,利用近红外光谱技术研究了复方氯丙那林和复方对乙酰氨基酚两种模型制剂有效组分的快速同时测定,建立了它们的多元校正模型,并以传统的稳健方法偏最小二乘回归(PLSR)为基础考察了PCA-SVMR方法对于小样本药物体系的拟合能力、预测能力和模型稳定性。研究表明,PLSR的预测能力必须以强拟合能力为前提,PCA-SVMR则没有这样的要求,使前者对校正样本的依赖性远强于后者,从而在小样本药物体系中前者的稳定性大大弱于后者,该两种药物制剂的PCA-SVMR多元校正模型的测定准确度总体上优于PLSR。

Based on the method of principal component analysis-support vector machine regression （PCA-SVMR）, the fast and simultaneous determination of the effective components in two model pharmaceutical preparations, compound clorprenaline and compound paracetamol, using near infrared diffuse reflectance spectroscopy was studied and their multivariate calibration models were established. The fitting capability, prediction capability and the robustness of the PCA-SVMR models built using small sample systems （i.e. their calibration sets containing a little samples） were compared with traditional robust method partial least squares regression （PLSR）. The results showed that the prediction capabilities of PLSR models depend on the precondition of their excellent fitting capabilities,but the PCA-SVMR models do not. Therefore, the dependent degree of PLSR models on the calibration samples is more than the PCASVMR models, and for the model robustness in the small sample systems of the studied drugs, the former is weaker than the latter, which makes the accuracies of their PCA-SVMR models be superior to the PLSR models on the whole for the two kinds of the researched drugs.