Optimizing Concrete Compressive Strength Prediction: A Comparative Study of Machine Learning Algorithms

This study investigates the predictive capabilities of LBGM, CATB, GBR, ADAB, and XGB models for concrete compressive strength prediction. Through an evaluation of default hyperparameters and comprehensive metrics, including R2, MSE, RMSE, MAE, RMSLE, and MAPE, nuanced insights into each algorithm’s performance are obtained. ADAB emerges as a standout performer, displaying lower error rates across multiple metrics, suggesting AdaBoost’s suitability for concrete compressive strength prediction. The analysis of a self-prepared dataset reveals significant variations across different mixture combinations and curing times. This research not only establishes a benchmark for current practices but also provides avenues for future research, including hyperparameter tuning to optimize model performance further. The practical implications for concrete engineering are significant, guiding material composition decisions and contributing to the development of more durable and resilient structures. As the intersection of machine learning and concrete engineering progresses, this study lays a foundation for tailored approaches to address the specific challenges of predicting concrete compressive strength.