Data and Python Code for Implementing Boosting Machines to Predict Local Buckling Strength of CFS Channels with Staggered Web Perforations

This repository contains the dataset and reproducible Python notebook used to benchmark design-code (DSM) calculations against machine-learning (ML) models for cold-formed steel channel members with and without staggered slots. The data files provide the master experimental dataset and DSM-based comparison sheets for the two specimen families (slotted and solid). The Jupyter notebook loads these files, performs preprocessing, trains/evaluates the ML models, and produces figures/tables comparing model accuracy to DSM.

File inventory

Data.xlsx — Master dataset used for ML training/validation. Contains geometric/material descriptors and the measured target response (bending capacity
𝑀
M, units as in the paper).

data_DSM_slots.xlsx — DSM benchmark calculations for slotted specimens (inputs + DSM-predicted capacities).

data_DSM_slot_1.xlsx — A curated subset of the slotted dataset (used for ablation/cross-checks or quick runs).

data_DSM_solids.xlsx — DSM benchmark calculations for solid (unperforated) specimens.

data_DSM_solid_1.xlsx — A curated subset of the solid dataset (used for ablation/cross-checks or quick runs).

ML paper-Copy1.ipynb — Reproducible Jupyter notebook: data loading, preprocessing, model training (e.g., XGBoost and baselines), metrics, and plots; includes DSM vs ML comparisons.

Notes

Typical feature columns include geometric descriptors and material properties (e.g., web/flange/lip dimensions, thickness, slot geometry descriptors, yield stress 𝑓𝑦 and the measured capacity 𝑀. Exact column names/units are documented in the notebook cells.

All spreadsheets are UTF-8 .xlsx files; the notebook targets Python 3.9+