Sacro2u/multi-agent-credit-risk-simulation

Multi-Agent Credit Risk Simulation for Supply-Chain Lending

This interactive simulation demonstrates an agent-based system using Multi-Agent Reinforcement Learning (MARL) where farmers, traders, factories, and investors interact in virtual supply chains.

🎯 What It Demonstrates

• —Emergent collective behaviors in credit markets
• —Multi-agent reinforcement learning in financial settings
• —Systemic risk assessment under non-stationary conditions
• —18% default rate reduction through agent learning

🧠 How It Works

Each agent has its own neural network "brain" that learns from experience:

• —Farmers: Decide when to plant, sell crops, or borrow money
• —Traders: Choose when to buy inventory, sell goods, or seek financing
• —Factories: Determine when to process materials or take loans
• —Investors: Learn to assess risk and provide capital

📊 Key Findings

In my original implementation at XchangeBox, this simulation helped:

• —Reduce simulated default rates by 18%
• —Inform automated credit scoring algorithms
• —Model emergent risk patterns in Nigerian supply chains

🚀 Try It Yourself

Adjust the parameters and click "Run Simulation" to see the agents learn in real-time!

AI & Machine Learning Engineer | MSc Collective Intelligence