An quality evaluation method based on three-dimensional integration and machine learning: Advanced data processing

Abstract

This study presents an innovative approach for the quality evaluation of traditional Chinese medicine (TCM) by integrating three-dimensional (3D) data processing with machine learning, aimed at enhancing the efficiency and accuracy of HPLC-DAD data analysis. Through 3D data integration, multi-dimensional signals from the time and wavelength domains are transformed into two-dimensional data, simplifying the analytical process while ensuring precise quantification of component contents. Building on this foundation, dynamic time warping (DTW) and correlation optimized warping (COW) algorithms were applied to effectively resolve retention time drift across different sample batches, achieving both global and local alignment of chromatographic peak shapes. A Binary Evaluation System (BES), incorporating macro qualitative similarity (S_m) and macro quantitative similarity (P_m), was employed to provide a comprehensive assessment of the quality of TCM samples. Additionally, machine learning models such as Multiple Linear Regression (MLR), Decision Tree Regression (DTR), and Random Forest Regression (RFR) were introduced to further improve the automation and accuracy of the evaluation system. In the analysis of 20 Scutellaria baicalensis samples, the method demonstrated a prediction error range of ±0.2 % for Baicalin content. This approach not only enhances data processing efficiency and reduces experimental resource consumption but also provides a robust theoretical and technical foundation for TCM quality assessment. Ultimately, the results of this study confirm the broad applicability of 3D integration and machine learning in TCM quality control, offering innovative technical support for the modernization of TCM quality evaluation systems.