lGikundiro/FXGuard-AI

Short-Term Exchange-Rate Depreciation Risk Classification and Decision Support aiming to help Rwanda-Based Importers Using Machine Learning. This project is a prototype focusing on three currencies: USD, EUR, and KES. # FXGuard AI — Multi-currency Exchange Rate Risk Forecasting FXGuard AI is a machine learning-driven, progressive web application for classifying short-term foreign-currency/RWF depreciation risk and estimating payment-planning scenarios for Rwanda-based importers. It does not stop depreciation, hedge a currency position, or provide financial advice. The prototype supports **USD, EUR, and KES against RWF**. Users select an invoice payment date from tomorrow through 100 days ahead; each currency has one horizon-aware classifier trained on every integer period from 1 to 100 calendar days. Hosted version: fxguard-ai-web.onrender.com API and fallback interface: fxguard-ai.onrender.com ## What the project includes - Official BNR buying, average, and selling rates imported from Excel exports - More than four years of USD/RWF, EUR/RWF, and KES/RWF history - Three horizon-aware classifiers covering payment dates 1–100 days ahead - FastAPI backend - Web interface frontend - Decision-support output for importers - Optional phone or email accounts using one-time verification codes - User-owned cloud history, account export, check deletion, and account deletion - Plain-language Privacy Notice and Terms of Use - Feedback form for usability testing - Training script to retrain the models ## Project structure ```text FXGuard_AI_Project/ backend/ app/main.py # FastAPI backend models/ # per-currency risk models and metadata data/ raw/ # uploaded BNR histories for USD, EUR, and KES processed/ # clean, feature, and model-ready datasets data_dictionary.csv frontend/index.html # web UI markup frontend/styles.css # frontend styling frontend/app.js # frontend behavior and API calls frontend/accounts.js # account, consent, and cloud-history behavior supabase/migrations/ # PostgreSQL tables and row-level access policies scripts/s …