FinRock

Reinforcement Learning package for Finance

Environment Structure:

Install requirements:

pip install -r requirements.txt
pip install pygame
pip install .

Create sinusoid data:

python bin/create_sinusoid_data.py

Train RL (PPO) agent on discrete actions:

experiments/training_ppo_sinusoid.py

Test trained agent (Change path to the saved model):

experiments/testing_ppo_sinusoid.py

Environment Render:

Links to YouTube videos:

• Introduction to FinRock package
• Complete Trading Simulation Backbone
• Training RL agent on Sinusoid data
• Included metrics and indicators into environment

TODO:

• [ ] Train model on continuous actions (control allocation percentage)
• [ ] Add more indicators
• [ ] Add more metrics
• [ ] Add more reward functions
• [ ] Add more scalers
• [ ] Train RL agent on real data
• [ ] Add more RL algorithms
• [ ] Refactor rendering, maybe move to browser?