llamidiy/miva-fraud-detection-dss

Machine Learning-Based Fraud Detection and Decision Support System for Digital Banking Transactions in Nigeria # Fraud Detection Decision Support System — Deployment Package This is a **staging copy**, prepared by Sprint 6.7.3, containing only the files the deployed Streamlit application genuinely needs at runtime — traced through actual imports and file reads, not copied wholesale from the research repository. See **`DEPLOYMENT_MANIFEST.md`** for the complete breakdown (entrypoint, dependencies, runtime files, exclusions, size, Linux compatibility, smoke-test results, and next steps), and **`DEPLOYMENT_CHECKSUMS.txt`** for SHA-256 verification of every model artifact. ## Quick Facts - **Entrypoint:** `app/app.py` - **Total size:** ~10 MB - **Model:** XGBoost only (`models/xgboost/model.joblib` + `encoder.joblib`) - **Excluded:** the 471 MB raw dataset and 946 MB engineered dataset — neither is used at runtime ## Running Locally ``` pip install -r requirements.txt streamlit run app/app.py ``` This directory is not yet a git repository and has not been pushed anywhere — see `DEPLOYMENT_MANIFEST.md` section K for the exact next steps.