DSSA-Optimized Hybrid Predictive Models for Estimating Concrete Compressive Strength and Slump

Concrete compressive strength (CS) and slump (SL) are two important properties that directly affect structural performance and workability, making accurate prediction essential for efficient mix design and quality control. This study develops Differential Squirrel Search Algorithm (DSSA)-optimized hybrid machine learning models, including ETDS (Extra Trees optimized by DSSA), DTDS, (Decision Tree optimized by DSSA), and NBDS (Naive Bayes optimized by DSSA), by integrating the Differential Squirrel Search Algorithm (DSSA) with Extra Trees, Decision Tree, and Naive Bayes regressors to improve hyperparameter optimization and predictive robustness. Existing studies do not report the integration of DSSA with machine-learning regressors for concurrent prediction of compressive strength and slump, establishing the key novelty of the proposed framework. The models were trained and validated on 189 experimentally measured concrete mixtures incorporating influential parameters such as water-to-binder ratio (W/B), sand-to-aggregate ratio (S/A), fine aggregate, air entrainment, silica fume, and superplasticizer (SP). The ETDS model achieved the highest predictive performance, with R² = 0.984 and RMSE = 3.468 for CS, and R² = 0.970 and RMSE = 5.650 for SL, outperforming both the standalone models and the other hybrid variants. SHAP analysis showed W/B, S/A, and SP as the most influential parameters, improving model transparency. DSSA-enabled hybrid ML models provide a reliable approach to optimizing concrete mixes, reducing trial-and-error, and supporting practical civil engineering decisions.