WATERMELON: Multi-Agent Reinforcement Learning Based Algorithmic Stock Trading System with GUI Application
1. WATERMELON
This repository is to introduce a multi-agent stock trading algorithm with a jointed policy distribution trained under strategy of deep reinforcement learning. The source code includes stock trading environment built with GYM API, various reinforcement learning algorithms (A2C, DDPG, PPO, ... etc.), and GUI application. As to RL algorithms, the numerical parts are largely built with numpy, but deep learning and linear algebra are to be accelerated and parallelized by using Tensorflow/Pytorch and JAX.
Unfortunately, the source code is currently under maintenance but of which alpha version is going to be released as soon as possible!
2. Requirements
requirements = {
tensorflow>=2.6.0
numpy~=1.19.5,
scipy~=1.4.0,
jax,
gym
pandas,
yfinance,
stockstats,
}axuiliary_requirements = {
opencv-python,
pyqt5,
matplotlib,
}3. Installation
This would be updated soon!
4. Usage
This would be updated soon!
5. Contribution and Issue
This would be updated soon!
6. Resource
This would be updated soon!
7. Reference
[1] Yang, Hongyang and Liu, Xiao-Yang and Zhong, Shan and Walid, Anwar, Deep Reinforcement Learning for Automated Stock Trading: An Ensemble Strategy (September 11, 2020). Available at SSRN: https://ssrn.com/abstract=3690996 or http://dx.doi.org/10.2139/ssrn.3690996