Comparative Analysis of Machine Learning Techniques for Concrete Compressive Strength Prediction

The compressive strength of concrete is of utmost importance in ensuring the structural stability and integrity of construction projects. With increasing importance placed on this aspect, there is a demand for intelligent and efficient techniques to accurately predict and evaluate compressive strength. Machine learning methods (ML) present a promising approach to developing accurate models for assessing the compressive strength of concrete. By leveraging the power of data analysis and pattern recognition, these techniques enable the construction industry to obtain reliable and robust concrete for a wide range of applications. ML models can identify the key factors influencing compressive strength, allowing for more informed decision-making, and ensuring the construction of structurally sound and durable projects. This article explores the assessment and comparison of four machine learning (ML) methods, namely Gradient Boosting, Categorical Gradient Boosting CatBoost, Random Forest (RF), and extreme gradient boosting (XGBoost) regressors, for predicting concrete compressive strength. Among these techniques, the CatBoost Regressor demonstrated superior performance in terms of accuracy, R-squared coefficient of determination, and Root Mean Square Error RMSE, surpassing the other three methods.