Artificial bee colony optimized random forest model for prediction of fly ash concrete compressive strength

Concrete compressive strength (CS) is a vital parameter in structural engineering, yet traditional evaluation methods are time-consuming, destructive, and resource intensive. This study presents a data-driven, computational approach for predicting the CS of fly ash concrete (FAC) using a hybrid Artificial Bee Colony (ABC)-optimized Random Forest (RF) model. The experimental dataset consists of concrete mix designs with varying proportions of fly ash, steel fibers, and water-to-binder ratios, tested over multiple curing ages. Among the regression models, RF algorithm demonstrating the highest predictive accuracy was further improved by tuning its hyperparameters through ABC optimization. The proposed ABC-RF model achieved an R² of 0.95 and outperformed several state-of-the-art deep learning models. The model also identified that a 30 % fly ash replacement yields optimal compressive strength, a finding further supported by SEM analysis, which showed dense C-S-H matrix formation at this level. The study demonstrates the effectiveness of bio-inspired optimization in civil material modelling and contributes toward sustainable construction practices. The method involves:•Experimental dataset generated from 168 fly ash concrete mix designs with varied input proportions.•Five regression models trained and evaluated using RMSE, MAE, and R².•Hyperparameters of the RF model optimized using the Artificial Bee Colony algorithm.