Data and code for black-box-guided symbolic regression of headed stud shear capacity in ultra-high-performance concrete (UHPC)

This repository provides the database, bibliographic sources, and source code used in a study on black-box-guided symbolic regression for predicting the shear capacity of headed studs embedded in ultra-high-performance concrete (UHPC).
The database contains 203 samples collected from published push-out tests and finite element analyses, including raw geometric, material, and arrangement parameters, together with the corresponding literature sources. Processed files are also included, containing the backbone term, normalized correction target, and dimensionless correction descriptors constructed for hierarchical modeling.
The code covers black-box machine learning, hyperparameter search, SHAP interpretation, symbolic regression with PySR, and candidate-equation ranking based on predictive performance, simplicity, and physical plausibility.
This repository is intended to support result reproduction, independent benchmarking, and reuse of the proposed workflow in related structural engineering applications.