Machine Learning-Based Prediction of Compressive Strength and Sustainability Performance of Cactus Fiber-Reinforced Concrete

The growing demand for sustainable construction materials has encouraged the adoption of natural fibers as alternatives to synthetic reinforcements in concrete. Cactus fiber (CF) has emerged as a promising reinforcement due to its renewable nature, favorable mechanical properties, and low environmental impact. However, predicting the nonlinear compressive strength behavior of cactus fiber-reinforced concrete (CFRC) remains challenging using conventional empirical approaches. This study applied four advanced machine learning algorithms – Support Vector Regression – Multilayer Perceptron (SVR-MLP) ensemble, Light Gradient Boosting Machine (LightGBM), Extreme Gradient Boosting (XGBoost), and Extreme Learning Machine (ELM) – to predict CFRC compressive strength. A comprehensive dataset was compiled from published literature incorporating key variables such as cement content, aggregate proportions, fiber dosage, water – binder ratio, and curing age. Model evaluation results showed that XGBoost achieved superior predictive performance with the highest accuracy (R2 = 0.9720) and lowest prediction errors compared to other models. The developed machine learning framework enables reliable strength prediction and optimization of fiber dosage under controlled mix conditions. Findings confirm cactus fiber as a viable low-carbon reinforcement material and demonstrate the effectiveness of artificial intelligence in supporting sustainable, data-driven concrete mix design and enhanced performance reliability.