This project offers a stock market trading simulation environment built with OpenAI Gym, targeting researchers and practitioners interested in applying deep reinforcement learning (DRL) to financial markets. It provides a foundational framework for developing and testing DRL agents, enabling users to experiment with different algorithms and model architectures for potentially improved trading strategies.

How It Works

The environment simulates stock trading using daily closing prices as input, allowing for custom data integration. It implements Deep Q-learning (DQN) and Policy Gradient (PG) algorithms, drawing inspiration from established DRL research. The core advantage lies in its flexibility as an open environment, encouraging users to modify and enhance the agent architectures and feature engineering for novel solutions.

Quick Start & Requirements

• Install via pip: pip install numpy h5py keras gym (backend like Theano or TensorFlow required).
• Requires Python 2.7+ or higher.
• Training data (e.g., CSV of closing prices) is needed. The provided samples are for Korean stocks.
• Official documentation or demo links are not explicitly provided in the README.

Highlighted Details

• Implements both Deep Q-learning and Policy Gradient algorithms.
• Built as a general-purpose OpenAI Gym environment for stock trading.
• Encourages user modification of model architecture and features.
• Training curves for a Policy Gradient agent on KOSPI stocks are shown.

Maintenance & Community

No specific information on contributors, sponsorships, community channels (Discord/Slack), or roadmaps is available in the README.

Licensing & Compatibility

The README does not explicitly state a license. Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

The provided neural network architecture is noted as too small for general use and may underfit with larger datasets; users are expected to redesign it. The project primarily uses Korean stock data, requiring users to source their own data for different markets.