Genetic Algorithm Optimized Bagging Ensemble Learning Approach for Estimating Concrete Compressive Strength From Recycled Concrete Aggregates

Prediction of compressive strength of concrete is a tough and crucial task in civil engineering because the material composition of concrete may vary along with external factors causing its performance to be different. Due to non-linear relationships that exist between input variables in traditional regression models, especially cement content, water-cement ratio, and properties of aggregates, they hardly capture such relationships accurately. Material property variations, curing conditions, and environmental factors further add complexity to the task. The traditional models often fail to capture these complexities, thus showing a limited level of predictability and reliability. A hybrid optimization framework is used, which combines Genetic Algorithms (GA) and Random Forest Regressor (RFR) to optimize key model hyper parameters, such as the number of estimators and the criterion. <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$G A$</tex> is efficient in exploring the hyper parameter space and enables the RFR to model complex interactions in the dataset with high precision. Experimental results show that the proposed model significantly improves the prediction accuracy with a low MSE and high correlation between predicted and actual compressive strengths by using a concrete mix composition dataset.