Data and code for: SHAP-Based Interpretability of Machine Learning Models for the Shear Capacity of Corrugated-Web Steel Girders
Description
Dataset, EN 1993-1-5 Annex D implementation, trained CatBoost model, SHAP attributions and analysis scripts supporting the article "SHAP-Based Interpretability of Machine Learning Models for the Shear Capacity of Corrugated-Web Steel Girders: Insights for Design-Code Calibration". The data comprise 181 experimentally tested steel girders with trapezoidal corrugated webs, a secondary compilation derived from the 206-specimen database of Shrif et al. (2024, Heliyon 10, e35778) by removing 25 specimens whose corrugation geometry falls outside the trapezoidal idealisation of EN 1993-1-5 Annex D. The deposit reproduces every numerical result, table and figure in the article, including the grouped SHAP attribution, the subgroup analysis of code-unconservative specimens, and the cross-validated trial recalibration of Annex D. The included EN 1993-1-5:2006 Annex D implementation has been verified expression by expression against the published standard, equations (D.4) to (D.10). The same 181 specimens underlie a companion study on tri-code comparison, deposited separately at https://doi.org/10.17632/nsm2gxj88n.1. See README.md for full provenance, a file manifest, and reproduction instructions.
Files
Institutions
- Rajamangala University of TechnologyPathum Thani, Thanyaburi