Effective carbon footprint assessment strategy in fly ash geopolymer concrete based on adaptive boosting learning techniques.

In light of the growing need to mitigate climate change impacts, this study presents an innovative methodology combining ensemble machine learning with experimental data to accurately predict the carbon dioxide footprint (CO-FP) of fly ash geopolymer concrete. The approach employs adaptive boosting to enhance decision tree regression (DTR) and support vector regression (SVR), resulting in a robust predictive framework. The models used key material features, including fly ash concentration, fine and coarse aggregates, superplasticizer, curing temperature, and alkali activator levels. These features were tested across three configurations (Combo-1, Combo-2, Combo-3) to determine optimal predictor combinations, with Combo-3 consistently yielding the highest predictive accuracy. The performance of the developed models was assessed based on standard metric indicators like mean absolute error (MAE), root mean square error (RMSE), Nash Sutcliffe efficiency (NSE), and correlation coefficient between the predicted and actual CO-FP. Results demonstrated that the Adaboost-DTR model with Combo-3 configuration achieved the best performance metrics during testing (CC = 0.9665; NSE = 0.9343), outperforming both standalone and other ensemble models. The findings underscore the value of feature selection and boosting techniques in accurately estimating CO emissions for sustainable construction applications. This research offers remarkable benefits for policymakers and industry stakeholders aiming to optimize concrete compositions for environmental sustainability. The results support future integration with IoT systems to enable real-time CO monitoring in construction materials. Finally, this study establishes a foundation for developing efficient CO-FP emission management tools.