frankacquah-blip/banking-cybersecurity-ews-xgboost-shap: v1.0 — SDT-Weighted XGBoost pipeline (MSc thesis submission)

Initial archived release of the modelling pipeline for "SDT-Weighted XGBoost for Explainable Early-Warning Prediction of At-Risk Staff in Banking Cybersecurity Onboarding: A Single-Institution Study in Ghana" (MSc Cybersecurity thesis, KNUST). Contains the six pipeline scripts that produced every result reported in the Methods and Results sections: 01_preprocessing.py — censoring filter, leakage-safe imputation/encoding, SDT composite scoring 02_modelling.py — Optuna hyperparameter tuning, baseline and SDT-weighted asymmetric-loss model training 03_smote.py — SMOTE resampling (training partition only) and retraining 04_shap.py — SHAP global/local explanations and feature pruning 05_evaluation.py — out-of-fold threshold selection, cross-validation, McNemar's test, bootstrap CI, timing 06_multiseed_sensitivity.py — 5-seed retraining stability check The real participant dataset is not included, due to pending institutional ethics approval (HuSSREC, KNUST); a data-use agreement is required for access, per the Ethical Statement in README.md.