Advanced machine learning-driven prediction in construction materials: comparative analysis and software implementation for fiber-reinforced graphene nano-engineered concrete

Abstract Fiber Reinforced Graphene Nano-Engineered Concrete (FRGNCC) is a high-performance, durable, and sustainable material. FRGNCC enhances strength, crack resistance, and service life while reducing CO 2 emissions and overall production cost, making it ideal for modern resilient structures. FRGNCC compressive strength, production cost, and CO 2 emissions are predicted using advanced hybrid machine-learning (ML) algorithms. Five hybrid ML models were developed using 11 input features: (1) ANN-PSO, (2) KNN-PSO, (3) RF-PSO, (4) SVR-GB, and (5) XGB-PSO. A comprehensive dataset compiled from peer-reviewed sources was used, and the models were evaluated using regression metrics. The XGB-PSO model is the most accurate and reliable in predicting the compressive strength of concrete, achieving a very high coefficient of determination (R 2 = 0.981 in the test data and = 0.991 in the training), with the lowest error indices (MAE = 0.604, RMSE = 0.748). Other models performed relatively poorly. RF-PSO came in second with good predictive ability but an error increase of approximately 3 % compared to XGB-PSO, followed by SVR-GB with a slight bias toward overprediction (+5 % error). KNN-PSO showed greater sensitivity at low and high resistance values, with an error increase of approximately 7 %, while ANN-PSO was the least accurate of the tested models, with an error increase of approximately 10 %. The developed hybrid models demonstrated outstanding predictive capability across mechanical, environmental, and economic aspects of FRGNCCs, where XGB-PSO consistently outperformed all other models, achieving near-perfect accuracy in tensile, flexural, and compressive strength prediction (R 2 up to 0.999) with minimal errors, while sensitivity analysis confirmed that cement content and curing age are the most influential factors in strength. Furthermore, the models accurately predicted production cost and CO 2 emissions (R 2 > 0.96) with very low relative errors (2–3%), highlighting their reliability as robust tools for multi-objective optimization in sustainable concrete design. The ML framework is designed for easy integration into a GUI, enabling engineers and researchers to efficiently estimate mechanical properties, cost, and CO 2 emissions of FRGNCC.