An OpenAI Gym environment for stock market trading simulation with Deep Q-learning and Policy Gradient implementations.
Stock Market Reinforcement Learning is an open-source project that provides an OpenAI Gym environment for simulating stock market trading using reinforcement learning algorithms. It includes implementations of Deep Q-learning and Policy Gradient methods to train agents on historical stock price data, allowing developers to experiment with algorithmic trading strategies. The project focuses on creating a flexible research platform rather than offering a complete trading solution.
Researchers and developers interested in applying reinforcement learning to financial markets, particularly those exploring algorithmic trading strategies using historical stock data.
It offers a ready-to-use, customizable environment for experimenting with reinforcement learning in stock trading, saving time compared to building such simulations from scratch. The open design encourages modification and improvement of models, making it ideal for academic and experimental projects.
This project provides a stock market environment using OpenGym with Deep Q-learning and Policy Gradient.
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Provides an OpenAI Gym-compatible environment for stock trading simulation, allowing seamless integration with various reinforcement learning algorithms, as stated in the README's overview.
Supports any stock data source beyond the included Google Finance and Korean samples, enabling users to apply their own datasets for training, as highlighted in the usage notes.
Emphasizes experimentation over a fixed solution, encouraging users to modify neural network architectures and features to develop better strategies, per the project philosophy.
Includes Deep Q-learning and Policy Gradient methods based on established research, serving as a practical foundation for algorithmic trading experiments, referenced from Karpathy's post.
The provided neural network is too small and may under-fit for large datasets, requiring users to redesign it for better performance, as admitted in the README's usage section.
Relies on Python 2.7 and older Keras backends like Theano, which are deprecated and may cause compatibility issues with modern libraries, based on the requirements list.
README is minimal with incomplete features, such as a to-do list for overfitting tests and Policy Gradient interface improvements, indicating a lack of polish.