Hybrid metaheuristic optimized Catboost models for construction cost estimation of concrete solid slabs

Accurate construction cost prediction is essential for effective project planning and resource allocation, particularly in the competitive construction industry. This study introduces an advanced approach to predicting the costs of concrete solid slabs by combining the Categorical Boosting (CatBoost) model with three hybrid metaheuristic optimization models: Phasor Particle Swarm Optimization (PPSO-CatBoost), Dwarf Mongoose Optimization (DMO-CatBoost), and Atom Search Optimization (ASO-CatBoost). These hybrid models were designed to optimize critical hyperparameters, including depth, learning rate, and iterations, and were benchmarked against a standalone CatBoost model using various performance metrics, such as residual error cumulative distribution (REC) curves, scatter plots, violin plots, and quantitative measures. The results reveal that the hybrid models consistently outperform the standalone CatBoost model, with ASO-CatBoost achieving the best overall performance with determination coefficient (R²) of 0.981 and root-mean-squared-error (RMSE) of 1.222 $/m². The ASO-CatBoost exhibited superior accuracy and generalization, characterized by minimal residual errors and close alignment with actual cost values during both training and testing phases. SHapley Additive exPlanations (SHAP) analysis identified the Tributary Area ($/m²) as the most influential variable, followed by Concrete ($/m³), underscoring the importance of these inputs in cost prediction. Additionally, a Python-based graphical user interface (GUI) was developed, enabling practical and user-friendly cost estimation in real-world applications.