What it does
VeighNa, published in the vnpy repository, is a Python development framework for quantitative trading systems. Its trader platform combines gateways to trading venues, strategy applications and a simple API for building custom modules. Gateways cover Chinese futures, options and stock markets as well as Interactive Brokers for overseas products, and the apps include a CTA strategy engine, a graphical backtester, spread, option and portfolio trading, algorithmic execution, paper trading and risk management. Version 4.0 added a vnpy.alpha module for machine-learning strategies. It covers feature engineering (including the Alpha 158 feature set from Microsoft's Qlib project), model templates for Lasso, LightGBM and MLP, strategy building from model signals, and a lab module that ties data management, training, signal generation and backtesting together. Demo notebooks show complete research workflows. The README, written in Chinese, also describes VeighNa Fusion, offered through partner futures firms, which helps turn a strategy description into code and connects to backtesting and parameter optimisation. Data can be stored in SQLite, MySQL, PostgreSQL and several time-series databases, and adapters for many data services are included. It needs Python 3.10 or newer, with install scripts for Windows, Ubuntu and macOS.
Capabilities
- Trader platform with gateways for Chinese futures, options and stocks and for Interactive Brokers
- Graphical CTA strategy backtester with parameter optimisation
- vnpy.alpha module for machine-learning workflows with Lasso, LightGBM and MLP models
- Algorithmic execution, spread, option and portfolio trading apps
- Database adapters and market-data service integrations
- Paper trading and risk management modules
When to use it
- Building and backtesting automated trading strategies in Python
- Researching machine-learning based multi-factor strategies on stock data
- Connecting a strategy to Chinese futures, options or securities accounts or to Interactive Brokers
