Prediction of Compressive Strength of Geopolymer Concrete via Machine Learning Techniques

In this study, a sum of 500 experimental test data was collected to estimate the compressive strength of Geopolymer concrete (GPC). Two prevalent machine learning approaches, artificial neural network (ANN) and random forest (RF), were established. Several input parameters, such as coarse and fine aggregates, alkaline solutions, alumino-silicate material, curing conditions, and concrete age, were taken into account. Meanwhile, the compressive strength of GPC was the output variable. The prediction results illustrated that the developed RF model displayed better performance in forecasting the compressive strength across both the training and testing sets. Conversely, the ANN model exhibited slightly lower accuracy in the testing set. Sensitivity analysis indicated that concrete age was the most important parameter significantly influencing the GPC’s compressive strength, while curing time showed the least importance among the variables.