This study investigates the mechanical performance of manufactured sand–based high-performance concrete (MSHPC) incorporating supplementary cementitious materials through an integrated experimental and data-driven approach. A total of 180 concrete mixes were prepared by varying manufactured sand replacement (0–100%), mineral admixture type and dosage (fly ash, silica fume, and metakaolin at 0–30%), and water–binder ratio (0.30–0.40). The results indicate that the optimal mix consisting of 60% manufactured sand and 10% metakaolin at a W/B ratio of 0.30 achieved the highest performance, at 28days compressive strength increasing from 68.79 MPa to 91.72MPa, (33.33% improvement), along with significant enhancements in split tensile and flexural strengths. The improved performance is attributed to enhanced particle packing, pore refinement, and interfacial transition zone densification. Mechanical performance decreased with increasing water–binder ratio, confirming the role of matrix densification in strength development. To enable predictive modeling, multiple machine learning models were developed, among which XGBoost demonstrated superior performance with R² values exceeding 0.99. SHAP-based explainable AI analysis revealed that curing age (for compressive strength), water–binder ratio, and manufactured sand content are the dominant governing parameters, while silica fume and metakaolin exhibited strong positive contributions toward strength enhancement. The proposed framework provides a data-driven and interpretable approach for optimizing sustainable high-performance concrete, reducing dependence on natural sand, and enabling efficient utilization of industrial by-products.
