Rapid quantification of active pharmaceutical ingredient for sugar-free Yangwei granules in commercial production using FT-NIR spectroscopy based on machine learning techniques

Rapid quantification methods for sugar-free Yangwei granules were developed based on near-infrared (NIR) spectroscopy combined with machine learning approaches as a quality control strategy for Chinese medicine granules (CMGs). Different machine learning approaches—i.e., interval partial least squares optimized by the genetic algorithm (GA-iPLS), the backpropagation artificial neural network (BP-ANN), and the particle swarm optimization-support vector machine (PSO-SVM)—were used to develop prediction models for three active pharmaceutical ingredients (APIs), namely, albiflorin, paeoniflorin, and benzoylpaeoniflorin. The partial least squares (PLS) algorithm was used for linear model calibration and comparison of the prediction performance of these developed models. The performance of the final models was assessed by the correlation coefficient (R), root mean square error of calibration set (RMSEC), and root mean square error of prediction set (RMSEP). All models performed well in model fitting and provided satisfactory prediction accuracy. The results indicate that the machine learning approaches are more stable, predictable, and suitable for CMGs when a high-accuracy analysis is required. In summary, NIR spectroscopy coupled with machine learning techniques is a suitable tool for the straightforward quantification of CMGs.