Credit Asset Demo – G20 TechSprint 2025

This project demonstrates a novel solution to Problem Statement 2: Consumer-consented and secure credit data portability, tailored for underbanked smallholder enterprises operating in informal or grey economies.

🔍 Concept

Traditional credit systems fail in environments where financial data is incomplete, informal, or inaccessible. This solution bypasses enterprise-level credit scoring by tying loan eligibility to individual asset performance — like a cassava farm — using real-time data and decentralized identity.

🎯 Use Case

A farmer in sub-Saharan Africa may have mixed business activities — some formal (e.g. cassava sales), some informal (e.g. egg sales at local markets). Rather than assessing their entire enterprise, this demo evaluates a specific productive asset using sensor data and AI models.

🧱 Architecture

• Frontend: A lightweight web dashboard where users connect a MetaMask wallet, consent to share data, and view AI-predicted yields
• Backend: Python Flask app serving sensor data, yield predictions (via a Random Forest model), and a federated comparison table
• Model: A dummy-trained ML model using soil and temperature to predict cassava yield
• Smart Contract: Ethereum-compatible contract that conditionally releases funds if yield exceeds threshold
• Federated Dataset: Simulated regional farm data for comparison, stored locally to demonstrate peer analysis
• User Flow Control: Tabs and buttons become active only in correct sequence (wallet > consent > yield > review > release)

💡 Features

• 🔐 Self-sovereign wallet connection (MetaMask)
• 📊 Sensor-driven yield forecasting
• 📈 Federated peer comparison (lender view)
• 🤝 Consent-first loan process
• ⚖️ Smart contract disbursement logic
• 🪙 CBDC/stablecoin-ready contract structure
• 📶 Real-time asset monitoring logic (expandable)

🏗️ Stack

• Frontend: HTML5, JavaScript, Chart.js, Web3.js
• Backend: Python, Flask, Pandas, Scikit-learn
• Smart Contract: Solidity (JSON ABI integrated in frontend)
• Blockchain: MetaMask testnet integration
• AI: Random Forest regression model (random_forest_model.joblib)

🚀 How to Run

1. Install dependencies

   pip install -r requirements.txt

2. Start Flask backend

   python flask_app.py

3. Open index.html in browser

4. Ensure MetaMask is connected to testnet

5. Optional: Deploy the smart contract separately to your testnet and update the contract address in the frontend if needed.

📁 Data Files

• cassava_farm_data.csv: Source for borrower predictions
• federated_farm_data.csv: Simulated data for lender comparison
• random_forest_model.joblib: AI model for predicting yields

📌 Notes

• Credit score is computed on the frontend using the average predicted yield:

  score = Math.min(Math.round(avgYield / 20), 100)

• FARM10 is synchronized across both borrower and federated datasets to ensure alignment.

📬 Contact

Built by Alis Grave Nil
For inquiries: alizgravenil@gmail.com

---

© 2025 – G20 TechSprint Demo – All rights reserved.