Supplementary Appendices for UHPC

Description:This repository contains the full supplementary material developed for the study on machine-learning-based prediction and symbolic regression modeling of shear capacity in Ultra-High-Performance Fiber-Reinforced Concrete (UHPFRC) beams. The appendices include: Appendix A:A complete list of experimental studies used to construct the UHPC/UHPFRC shear database, including bibliographic details for all non-prestressed and prestressed beam tests compiled from prior literature. These references were used to develop the machine-learning dataset and ensure transparency, reproducibility, and traceability of all data sources. Appendix B:A comprehensive comparison of feature selection techniques applied across multiple regression models (SVR, Random Forest, XGBoost, Linear Regression, etc.). Wrapper-based, embedded, permutation-importance, and genetic-algorithm (GA) methods are summarized, with selected features reported for both non-prestressed and prestressed beams. Evaluation metrics (R²) for each method are included to document their relative performance and justify the selection of the final feature subset for symbolic regression. Purpose:These appendices support the primary manuscript by providing detailed methodological transparency, enabling other researchers to reproduce the machine-learning workflow, verify dataset sources, and compare feature selection strategies. The material also serves as a reference for future studies on data-driven UHPC shear design and ML-based structural engineering research.