Construction of an interpretable model for airfoil aerodynamic characteristics based on the light gradient boosting machine—Shapley additive explanation framework

To achieve high-precision prediction and efficient optimization of airfoil aerodynamic performance, this study proposes a Light Gradient Boosting Machine (LightGBM)-based model for predicting airfoil lift (CL) and drag (CD) coefficients. The model's performance was compared with four classic machine learning algorithms (Random Forest, XGBoost, AdaBoost, and Gradient Boosting Machine), and its generalization ability was verified using a novel dataset. Combined with the Shapley value (SHAP) explainable model, feature analysis of CL and CD was conducted. Results show the LightGBM model exhibits optimal predictive performance: R2 values for CL are 0.996 (training set) and 0.9854 (test set), while those for CD are 0.9996 and 0.9858, outperforming other algorithms. In contrast, AdaBoost suffers severe overfitting and is unsuitable for the task. SHAP analysis identifies core control parameters (c4, c3, c13 for CL; c4, c6, c7 for CD) and key interaction combinations (e.g., c13 × c14 for CL, c3 × c4 for CD), quantifying parameter contribution intensities and breaking the “black-box” limitation of traditional models. The SHAP-based optimization strategy reduces computational fluid dynamics (CFD) simulation blindness, shortening R&D cycles and computational costs. This study integrates high-precision prediction, interpretable analysis, and efficient optimization, providing a cost-effective technical path for rapid design and engineering application of high-performance airfoils.