Multi-Target Strength Prediction of Steel Fiber Reinforced Concrete Using XGBoost and SHAP-Based Interpretation

The paper presents a machine learning approach using the XGBoost model to perform multi-target prediction of the compressive, tensile, and flexural strengths of Steel Fiber Reinforced Concrete (SFRC). After training the model on experimental data, the SHAP (SHapley Additive exPlanations) technique is employed to analyze and interpret the influence of input features such as steel fiber content, cement dosage, aggregates, and admixtures on the predicted outcomes. SHAP quantifies the contribution of each feature to the model’s predictions, thereby offering deeper insights into the relationships between material components and the strength properties of SFRC. The results show that the combination of XGBoost and SHAP not only achieves high prediction accuracy but also provides strong interpretability, effectively supporting the design and optimization of SFRC materials.