Home  ›  Research  ›  A Study on Stock Trading Strategies Based on Multi-Agent Reinforcement Learning

This study leverages multi-agent reinforcement learning to simulate competitive and cooperative trading behaviours, uncovering emergent strategies that outperform single-agent benchmarks in backtested market environments.

DeepCox Technical Support

• RL Environment: Developed a custom OpenAI Gym environment with realistic market dynamics.
• Multi-Agent Algorithms: Implemented MAPPO, QMIX, and MADDPG for comparative analysis.
• Feature Engineering: Created technical indicators and sentiment embeddings from news and social media feeds.
• Backtesting Framework: Built a robust backtesting engine with walk-forward validation and risk metrics.