A novel composite machine learning model for the prediction of compressive strength of blended concrete

Blended cement, comprising clinker and supplementary cementitious materials (SCMs), such as fly ash, slag, and silica fume, forms blended concrete when combined with aggregates. This study introduced a novel ensemble approach, the LightGBM and ANN Fusion (LAF) model, to check the prediction of compressive strength in blended concrete using eight key input features, including cement content, slag, fly ash, water, and curing age, derived from 1030 samples. The LAF model synergistically combines predictions from LightGBM and ANN with the fusion weight (alpha) dynamically optimized using Support Vector Machine (SVM) analysis. This integration leverages the strengths of both models while mitigating their individual limitations. The LAF model exhibited superior performance, with an R² of 0.9514 during training and 0.8863 during testing, significantly surpassing standalone models such as LightGBM (R² = 0.9385) and ANN (R² = 0.9136). It achieved the lowest error rates, including an MSE of 15.7384 and RMSE of 3.9672 during training, underscoring its robust generalization capabilities. The novel utilization of SVM-tuned alpha for dynamic model adaptation reduces overfitting and enhances predictive accuracy compared to traditional ensemble and single models. Comparative analysis with existing ensemble techniques, including stacking and gradient-boosted ANN, further validated the superior accuracy and adaptability of the LAF model. This approach represents a significant advancement in predictive modeling, offering a scientifically sound and versatile framework for concrete strength prediction. Future research will refine the SVM tuning, expand the model’s application to other material properties, and explore real-time deployment, positioning the LAF model as a pioneering tool for smart construction materials.