VeighNa 4.0: AI-Powered Quantitative Trading Framework
VeighNa, a popular open-source quantitative trading system based on Python, has released version 4.0. This update introduces the vnpy.alpha module, designed for professional quantitative traders to develop, research, and execute AI-powered multi-factor machine learning (ML) strategies. The module aims to provide a one-stop solution for ML-based quantitative trading, integrating data processing, model training, and strategy development.
Key Features of vnpy.alpha Module
The vnpy.alpha module is inspired by the Qlib project and offers several key features:
- dataset: Optimized for ML algorithm training, it supports efficient batch feature calculation and processing. It includes a built-in factor expression calculation engine for quick data generation and supports custom expression functions.
- Alpha 158: A set of stock market features derived from the Microsoft Qlib project, covering K-line patterns, price trends, and time-series volatility.
- model: Provides standardized ML model development templates, simplifying model construction and training. It supports seamless switching between different algorithms for performance comparison.
- strategy: Enables rapid construction of quantitative trading strategies based on ML signal prediction models, supporting both cross-sectional multi-asset and time-series single-asset strategies.
- lab: Integrates data management, model training, signal generation, and strategy backtesting into a complete workflow, with built-in visualization tools for intuitive evaluation.
Supported Machine Learning Models
The vnpy.alpha module integrates several mainstream machine learning algorithms, each with specific use cases:
| Model | Description | Use Case |
|---|---|---|
| Lasso | Classic Lasso regression model with L1 regularization for feature selection | Linear models with feature selection |
| LightGBM | Efficient gradient boosting decision tree optimized for large datasets | High-performance prediction on large datasets |
| MLP | Multi-layer perceptron neural network for complex non-linear relationships | Complex non-linear modeling |
Installation and Compatibility
VeighNa 4.0 is compatible with Python 3.10 and above, with Python 3.13 recommended. The framework supports various operating systems, including Windows 11, Windows Server 2022, and Ubuntu 22.04 LTS. Users can install VeighNa Studio, which includes the VeighNa framework and the VeighNa Station quantitative management platform, by downloading the release version and running the installation script.
Community and Support
VeighNa has a strong community presence, with official forums, a QQ group, and a WeChat group for user support. The project is hosted on GitHub, and contributions are welcome through the Pull Request process. Users are encouraged to follow coding standards, including using ruff for code style checks and mypy for static type checking.
Conclusion
The release of VeighNa 4.0 marks a significant step forward in AI-powered quantitative trading, providing professional traders with a robust and flexible framework for developing and executing ML-based strategies. The integration of the vnpy.alpha module, along with the support for multiple machine learning models, makes VeighNa a powerful tool for quantitative traders looking to leverage AI in their trading strategies.
Источник: Github ↗