Abstract
This dissertation presents innovative machine learning and metaheuristic optimization approaches to tackle key challenges in modern concrete technology and structural engineering. Three advanced predictive models were developed to improve accuracy, interpretability, and usability across different concrete systems. First, a Particle Swarm Optimization–Categorical Boosting (PSO-CatBoost) model was created to estimate the compressive strength of Carbon Fiber-Reinforced Polymer Confined-Concrete (CFRP-CC) specimens, using a large experimental dataset. SHapley Additive exPlanations (SHAP) and Permutation Feature Importance (PFI) analyses revealed key influencing factors, ensuring interpretability. Second, to predict the flexural strength of 3D-printed fiber-reinforced concrete (3DP-FRC), a novel Mountain Gazelle Optimization–Artificial Neural Network (MGO-ANN) model was proposed. Data augmentation using Kernel Density Estimation and Wasserstein GANs addressed data scarcity, and SHAP analysis identified dominant parameters. Third, a robust XGBoost model was developed to predict the punching shear strength of Fiber-Reinforced Polymer Reinforced Concrete (FRP-RC) slabs. Enhanced by kernel density estimation, the model achieved high accuracy, identifying effective slab depth as the most influential factor. For each model, a graphical user interface (GUI) was created to promote real-world use. These studies highlight the power of data-driven solutions to enhance predictive performance, inform design practices, and support sustainable construction.