建立测定克霉唑粉末药品的近红外漫反射光谱结合可移动窗口偏最小二乘法(MW-PLS)多元校正模型。选择有效的光谱预处理方法,并对窗口宽度,窗口移动的位置,以及隐变量数对模型参数进行优化,使用逼近度作为建模参数优化评价指标,最终得到测定克霉唑粉末药品的最佳模型。用该模型进行预测,校正集和验证集的预测值和真实值间的相关系数(Rc)分别为0.9558和0.9645,校正集的均方根误差(RMSEC)为0.0041,验证集均方根误差(RMSEP)为0.0043,表明该模型的稳健性、拟合度和预测能力都令人满意。
The quantitative analysis model of multivariate calibration method of the contents of Clotrimazole in Clotrimazole Pharmaceutical Powder was established with near-infrared diffuse reflectance spectroscopy combined with moving windows partial least square (MWPLS). The degree of the approaching (Da) was employed as criteria of selecting effective parameter of the building models. The MWPLS model was optimized by selecting the processing method, window width, the number of window and number of latent variables. The optimum model for Clotrimazole analysis was obtained with the above mentioned optimization. The correlation between predictive and actual values of the calibration samples and validation samples were 0.9558 and 0.9654, respectively. The root mean square error of the calibration samples and validation samples were 0.0041 and 0.0043, respectively. The results demonstrated that the stability, the fitting and predictive capability of established model were satisfactory.