A comparative analysis of tree-based machine learning algorithms for predicting the mechanical properties of fibre-reinforced GGBS geopolymer concrete

The advancement of strength prediction techniques for fibre-reinforced ground granulated blast furnace slag (GGBS)-based geopolymer concrete (GPC) is still in its early stages, especially compared to conventional GPC, primarily due to its complexity and the scarcity of available data. Therefore, this study evaluates five tree-based machine learning techniques, including decision tree, random forest, XGBoost (XGB), gradient boost and AdaBoost, to determine the optimal algorithm for forecasting the mechanical strength properties of fibre-reinforced GGBS-based GPC. The input parameters included the mix proportions, curing conditions and properties of fibres; while the output parameters were compressive strength, split tensile strength and flexural strength. The models were trained, tested and validated using a comprehensive dataset obtained from the published literature. The performance and accuracy of the models were compared based on statistical evaluation metrics such as the coefficient of determination (R2), mean squared error (MSE), root mean squared error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE). k-fold validation was performed to mitigate overfitting and the algorithms were fine-tuned for optimal performance. The results demonstrated that the XGBoost model outperformed other models, achieving high predictive accuracy with R2 values of 0.938, 0.957 and 0.909 for compressive, split tensile and flexural strength, respectively, along with significantly lower RMSE and MAE values. XGBoost continued to outperform other techniques and showed consistency of high accuracy in performance for the testing dataset even after k-fold cross-validation. The sensitivity analysis results indicate that the prediction of compressive strength in fibre-reinforced GGBS-based GPC was most significantly influenced by specimen age and curing temperature. However, the fibre volume and curing temperature strongly influenced split tensile and flexural strength.