论文部分内容阅读
采用紫外-可见漫反射光谱-支持向量回归法(UV-Vis DRS-SVR)建立了快速、无损测定西咪替丁片剂的定量方法。人工配制58个西咪替丁片剂粉末样,分为校正集和预测集。通过光谱预处理、异常值剔除和调整RBF核函数参数g、正则化系数C和不敏感损失函数p来优化模型,最终采用原始光谱,在220.17~980.80 nm,参数g=0.02,C=20,p=0.03的条件下,以支持向量回归法(SVR)建立模型,校正集和预测集的决定系数(R2)分别为0.978 2,0.987 1,对5个盲样和15个批次市售西咪替丁片剂的预测均方根差(RMSEP)分别为0.036 0,0.044 8,比偏最小二乘法(PLS)所建模型的预测精度更高,效果更好。研究表明,UV-Vis DRS-SVR用于快速、非破坏性测定药物制剂中的活性成分的含量是可行的,有望用于制剂生产过程中的质量控制。
A rapid and non-destructive method for the determination of cimetidine tablets was established by UV-Vis DRS-SVR using ultraviolet-visible diffuse reflectance spectroscopy. Fifty-eight cimetidine tablets were prepared manually and divided into calibration sets and prediction sets. The model was optimized by spectral preprocessing, eliminating outliers and adjusting RBF kernel function parameter g, regularization coefficient C and insensitive loss function p. Finally, the original spectra were used at 220.17-980.80 nm with parameters of g = 0.02 and C = 20, p = 0.03, the support vector regression (SVR) model was established. The coefficients of determination (R2) for the calibration set and the prediction set were 0.978 and 2.0.987 1 respectively. For the five blind samples and 15 batches of commercial products The root mean square error (RMSEP) of mirtine tablets were 0.036 0 and 0.044 8 respectively, which was higher than that predicted by partial least square method (PLS). Studies have shown that the use of UV-Vis DRS-SVR for the rapid and non-destructive determination of the active ingredient content in pharmaceutical preparations is viable and is expected to be of use in the quality control of the formulation production process.