Personae
A repository implementing Deep Reinforcement Learning and Supervised Learning methods with a simulated financial market environment for quantitative trading.
What is Personae?
Personae is an open-source repository that implements Deep Reinforcement Learning and Supervised Learning algorithms specifically designed for quantitative trading. It provides a simulated financial market environment where users can train and test trading agents using historical stock and future data. The project serves as a research toolkit for developing and comparing machine learning-based trading strategies.
Quantitative researchers, algorithmic traders, and machine learning engineers interested in applying RL and SL to financial markets. It's particularly useful for those who want to experiment with paper implementations in a controlled trading simulation.
Personae offers a curated collection of implemented papers with a ready-to-use trading environment, reducing the time needed to set up experiments. Its Docker support ensures reproducibility, and the modular design allows easy extension with custom features and algorithms.
Overview
📈 Personae is a repo of implements and environment of Deep Reinforcement Learning & Supervised Learning for Quantitative Trading.
Use Cases
Best For
- Testing Deep Reinforcement Learning algorithms for automated trading strategies
- Comparing Supervised Learning models for stock price prediction
- Academic research on machine learning applications in finance
- Building a baseline quantitative trading research environment
- Learning how to implement trading simulations with gym-like interfaces
- Experimenting with paper reproductions in a financial context
Not Ideal For
- Teams building production-ready trading systems due to naive input features and day-frequency data limitations
- Developers needing modern ML frameworks, as it relies on TensorFlow 1.4 and Python 3.5
- Researchers requiring high-frequency data support, since the project acknowledges day frequency is insufficient
- Users seeking a fully-featured out-of-the-box solution, given the incomplete short sale implementation and complex MongoDB setup
Pros & Cons
Pros
Provides clean, reusable implementations of 4 RL and 3 SL methods from academic papers, such as DDPG and DA-RNN, with TensorFlow, as detailed in the Contents section.
Includes a custom gym-like simulation for stock and futures trading, enabling realistic strategy testing without real market risks, as described in the Environment section.
Offers pre-built Docker images with CUDA support, simplifying setup and ensuring consistent experimental conditions, highlighted in the How to Use section.
Comes with training results and visualizations on historical data, such as profit charts and price predictions, helping users understand model performance and baseline comparisons.
Cons
Relies on Python 3.5 and TensorFlow 1.4, which are no longer mainstream and may have security or compatibility issues, as indicated by the Platform and Python badges.
The README admits that input features are naive and day frequency is insufficient, requiring users to replace features for better results, as noted in the Attentions section.
Short sale functionality is still under implementation, and the project is marked as under reconstruction with a warning, indicating instability and potential breaking changes.
Requires MongoDB for data storage and involves multiple steps with Docker and spiders for data crawling, which can be time-consuming and error-prone, as detailed in the How to Use section.
Frequently Asked Questions
Related Projects
FinRLFinRL®: Financial Reinforcement Learning. 🔥
stockpredictionaiIn this noteboook I will create a complete process for predicting stock price movements. Follow along and we will achieve some pretty good results. For that purpose we will use a Generative Adversarial Network (GAN) with LSTM, a type of Recurrent Neural Network, as generator, and a Convolutional Neural Network, CNN, as a discriminator. We use LSTM for the obvious reason that we are trying to predict time series data. Why we use GAN and specifically CNN as a discriminator? That is a good question: there are special sections on that later.
Ensemble-StrategyFinRL-X: An AI-Native Modular Infrastructure for Quantitative Trading
mlforecastScalable machine 🤖 learning for time series forecasting.
Found a gem we're missing?
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.