nshutielise/loan-approval-rwanda

# Loan Approval Prediction App (Rwanda-Specific Prototype) This interactive app simulates an intelligent loan screening system tailored to the **Rwandan banking sector**. It uses **machine learning** enhanced with **domain-inspired rules** to make explainable, risk-aware lending decisions. --- ## Purpose This is a **prototype simulation**, demonstrating how credit decisions can be: - Faster - Fairer - More transparent It is not a production tool yet — but can serve as a **foundation to build robust loan decisioning systems** when co-designed with domain experts in Rwanda’s financial services. --- ## Use Case Fit Designed for use by: - Commercial banks - MFIs and SACCOs - Digital lending startups - Credit scoring & analytics teams ### Additional Use: Bank clients can use this simulation to **pre-evaluate their own eligibility** before applying — reducing in-branch time and unnecessary paperwork. --- ## Lending Logic Based on Local Reality This app integrates **basic credit rules contextualized for Rwanda**: - Reject if **monthly income < RWF 120,000** - Reject if **loan exceeds 40% of annual income** - Reject if credit grade is **F or G** - Reject if employment length is **< 1 year** - Only accept loans with a predicted **repayment probability ? 65%** --- ## Prediction + Explainability Combines machine learning prediction with **visual explanations** using SHAP: - Predicts likelihood of full repayment - Displays a waterfall plot of feature contributions (SHAP) - Explains *why* an application is accepted or rejected --- ## Tech Stack | Component | Technology | |----------------|---------------------------------| | Frontend | Streamlit | | Model | XGBoost / Logistic Regression | | Preprocessing | Scikit-learn Pipelines | | Balancing | SMOTE (imbalanced-learn) | | Explainability | SHAP …