AutoGluon, SHAP, and GUI: A Groundbreaking Framework for Concrete Mechanical Strength Modeling, Explainability, and Accessibility

Accurate prediction of concrete mechanical strength is crucial for ensuring safe and dependable infrastructure design and construction. However, traditional empirical models often face challenges in precisely predicting mechanical strength due to the complex nonlinear relationship between concrete properties and strength. This study proposes an AutoGluon–Shapley additive explanations (AutoGluon‐SHAP) approach, which automatically predicts concrete mechanical strength while providing insightful interpretations of the results. Additionally, a graphical user interface (GUI) was developed to enhance global accessibility to the model. AutoGluon leverages powerful algorithms such as LightGBM, CatBoost, XGBoost, random forest, ensembles, and neural networks within its TabularPredictor to automate model selection. The models were trained and tested using a comprehensive dataset of 1760 concrete samples sourced from the published literature, encompassing diverse concrete types, cement brands, additives, target compressive strengths, and curing periods. Results show that the WeightedEnsemble_L2 predictive model demonstrated robust performance in estimating flexural, tensile, and compressive strengths. All the predictive models achieved high performance with R 2 of more than 0.98 for all the strengths’ properties. SHAP offered both global insights into the influence of the predictors on mechanical strength and local explanations for individual predictions, ensuring transparency and reliability. The developed GUI resulted in a globally accessible and reliable estimation of concrete mechanical strength.