albertameyaw/sp-xgboost-momo-fraud

SHAP-Pruned XGBoost for mobile money fraud detection in Ghana # SP-XGBoost: SHAP-Pruned XGBoost for Mobile Money Fraud Detection in Ghana Engineered XGBoost variant that prunes the bottom ~30% of features by mean absolute SHAP value, targeting equal detection performance at lower inference cost. Hybrid data: PaySim + ethically-cleared Ghanaian field records. ## Status Proposal stage — baseline and engineered-model notebooks to follow. ## Planned structure - data/ (no raw field data committed — see ethics) - notebooks/ (01_baseline.ipynb, 02_sp_xgboost.ipynb, 03_evaluation.ipynb) - requirements.txt ## Author Ameyaw Albert Adofo , Dept. of Computer Science, KNUST