Reinforcement Learning Trading System - V1.1.0

Clone the git repo, then install the requirements with pip

git clone https://github.com/shin2suka/RL-Trading.git
cd RL-Trading
pip install -r requirements.txt

Run the trading system

python backtest_main.py

Current algorithm: Deep Q learning

TO DO:

Urgent:

• Deploy CNN network
• Design a bette reward mechanism and the state space (NEED TO FIX profit_per_episode plot)
• Simulate several scenarios (e.g., up/down trend and mean reversion)
• Incorporate with performance engine to track stats

Non-urgent:

• Add more RL algorithms (policy gradient/PPO/A3C)
• Incorporate with dash visualization
• Build risk engine to control risk
• Build trading signals
• enable to hold multiple assets

Reference

Algorithm Trading using Q-Learning and Recurrent Reinforcement Learning

Learning to Trade via Direct Reinforcement

©2020 All Rights Reserved by Hongyi (Henry) Wu & Tianyu (Shin) Ren & Zhizu (Zale) Li