Extreme Gradient Boosting Tuned Algorithms for Estimating Fracture Toughness of Geopolymer Mortar Containing Ground Granulated Blast Furnace Slag

The correlation between mix design and fracture characteristics of ground granulated blast furnace slag (GGBFS) concrete is intricate and affected by several variables, including mixture composition. This study created novel and powerful models employing an extreme gradient boosting (XGB) algorithm to estimate the fracture toughness (FT) under mode-I (KIC), mode-II (KIIC), and mixed-mode, accomplished on centrally straight-notched Brazilian discs. The introduced data set was collected from literature, including 143 data rows of the FT of GGBFS mortar under different loading angle conditions. In order to time reduction and accuracy increment, the XGB linked with four cutting-edge optimizers as hyperparameter tuner, named Tasmanian devil optimization (TDO), Tunicate swarm algorithm (TSA), Pelican optimization algorithm (POA), and Jellyfish search optimization algorithm (JSOA). Moreover, the sensitivity analysis was performed on the input variables to determine the impact of each attribute on the targets (KIC and KIIC). The result of the created models depicted that all XGB-based models performed acceptably during the prediction outline with high accuracy and dependability, whereas the XGB optimized with POA reported the most appropriate performance compared to others capable of being utilized in practical efforts.