Predicting compressive strength of geopolymer concrete using machine learning models

Geopolymer concrete (GPC) has recently gained considerable attention for its potential to reduce carbon emissions and improve durability compared to traditional Portland cement-based concrete. Accurate prediction of the mechanical properties of GPC is essential for its successful application in construction projects. This study utilizes machine learning techniques such as statistical package for the social sciences (SPSS), artificial neural network (ANN), and python to predict the compressive strength of GPC. For prediction of compressive strength input parameters like fly ash quantity and fineness, solution-to-binder ratio, and excess water presence were used. Seventy percent of the dataset was used for training the models, while the remaining 30% was used for testing. The experimental results of GPC incorporating fly ash indicated similar or higher compressive strengths, especially at later ages, compared to standard concrete. The machine learning proposed models demonstrated high predictive accuracy, as indicated by a high correlation coefficient of 0.997 in Python, 0.981 in MATLAB ANN, and 0.940 in SPSS. The developed machine learning models showed high efficiency in predicting the compressive strength of geopolymer concrete. The findings highlight the potential of Python and ANN-based models for precise compressive strength prediction, facilitating GPC's practical application in structural engineering. This study concludes that machine learning models provide a powerful and efficient approach for forecasting the mechanical properties of GPC, supporting its broader adoption as a low-carbon construction material.