Enhanced classification of traditional Chinese medicine syndromes in primary lung cancer patients using stacking ensemble machine learning algorithms

This study aims to employ advanced machine learning techniques, particularly the Stacking ensemble algorithm, to accurately classify thirteen Traditional Chinese Medicine (TCM) syndromes in primary lung cancer patients. Data from 2,598 lung cancer cases were collected from nine hospitals nationwide, encompassing 125 clinical symptoms and signs. Data preprocessing included missing value imputation, feature selection, and standardization to ensure data quality and effective model training. Nine foundational models were trained: Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest (RF), Extra Trees, Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), Adaptive Boosting (AdaBoost), Gradient Boosting (GB), and Multi-Layer Perceptron (MLP). By selecting the top four performing basic models for the Stacking ensemble, classification performance was significantly enhanced. Experimental results demonstrated that the XGBoost fusion model achieved a micro-average Area Under the ROC Curve (AUC) of 0.996, outperforming single models and highlighting the efficiency and stability of ensemble learning in TCM syndrome diagnosis. This study confirms that the Stacking ensemble algorithm holds significant potential for complex multi-class classification tasks, offering a novel computational approach for TCM diagnostics.