Deep reinforcement learning framework for financial trading using price trailing, implemented with Keras-RL for Forex markets.
Trading-RL is a deep reinforcement learning framework for automated financial trading in Forex markets. It implements two different trading strategies using DQN agents from Keras-RL, including a novel price trailing approach presented at ICASSP 2019. The project provides a complete research environment for developing and testing reinforcement learning trading algorithms on currency exchange data.
Researchers and data scientists working on algorithmic trading, particularly those interested in applying deep reinforcement learning to financial markets. It's also suitable for machine learning practitioners exploring time-series forecasting and decision-making problems.
This project offers a research-ready implementation of cutting-edge reinforcement learning trading strategies with built-in comparison capabilities. Unlike generic RL frameworks, it provides domain-specific trading environments and visualization tools specifically designed for financial time-series data analysis.
Deep Reinforcement Learning for Financial Trading using Price Trailing @ ICASSP 2019
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Based on a peer-reviewed ICASSP 2019 paper, implementing a novel price trailing strategy that ensures research credibility and reproducibility.
Provides custom OpenAI Gym-compatible environments, a baseline model for comparison, and Bokeh visualization tools, offering a holistic setup for trading experiments as described in the README.
Allows CPU or GPU training with adjustable cores and memory allocation in dqn_agent.py, enabling parallel model execution and resource optimization.
Automatically saves model weights, architecture, training history, and performance metrics in organized directories, facilitating analysis and reproducibility post-training.
The README explicitly states the project is no longer active, with slow bug fixes and updates, making it unreliable for long-term or production use.
Relies on older Python versions (2.7/3.5) and deprecated libraries like keras-rl, which may cause compatibility issues with modern systems and hinder integration.
Requires manual preprocessing of Forex data, with notes on incomplete or poorly formatted datasets online, adding significant setup overhead for users.
Focused solely on Forex markets with specific environments, making it difficult to extend to other financial instruments without substantial code modifications.