A Comparative Predicting ML Model for Compressive Strength of Fly Ash/GGBFS Geopolymer Concrete

Abstract This research investigated the prediction of compressive strength in fly ash/GGBFS geopolymer concrete using three machine learning techniques: artificial neural network (ANN), multivariate adaptive regression splines (MARS), and MultiGene Genetic Programming (MGGP). The performance of these techniques was compared with traditional linear and nonlinear methods. Evaluation metrics such as correlation coefficient (R), root mean square error (RMSE), and mean absolute error (MAE) were used, along with Taylor diagrams, to conduct a thorough comparative analysis of the prediction models. Sensitivity and parametric analyses were performed to assess the contribution and effectiveness of individual input variables. The results indicated that MGGP outperformed the other models in predicting the compressive strength of fly ash/GGBFS geopolymer concrete. The study demonstrates the potential of predictive tools for concrete strength and emphasizes the importance of considering input parameters' impact on strength prediction. Experimental validation of the selected model further supported its accuracy.