|
Title |
Development of a Tree-Based Machine Learning Model for Predicting the Compressive Strength of Accelerated Carbonation-Cured Cementitious Composites
|
|
Authors |
조상환(Sanghwan Cho) ; 김민욱(Min Ook Kim) |
|
DOI |
https://doi.org/10.4334/JKCI.2026.38.1.091 |
|
Keywords |
가속 탄산화; 강도 예측; 트리 기반 알고리즘; SHAP 분석 accelerated carbonation; strength prediction; tree-based algorithms; SHAP analysis |
|
Abstract |
In this study, tree-based machine learning models were developed to quantitatively predict the compressive strength development of CO2-cured cementitious composites and to identify the key influencing factors based on experimental data collected from the literature. A dataset comprising 333 experimental results was compiled from 23 published studies, incorporating variables related to binder composition, mixture proportions, CO2 curing conditions, environmental parameters, and curing age. The predictive performance of Random Forest and Gradient Boosting algorithms was evaluated, and the results showed that CatBoost and XGBoost achieved high prediction accuracy and stable generalization performance on the test dataset. SHAP-based sensitivity analysis revealed that CO2 curing duration, coarse aggregate-to-binder ratio, and CO2 concentration were the dominant variables governing compressive strength development, exhibiting non-monotonic and nonlinear influence characteristics. This study provides an interpretable, data-driven framework for understanding compressive strength development under CO2 curing and offers a foundation for future multi-objective optimization studies that integrate carbon uptake efficiency and durability performance.
|