Project Overview

Machine Learning-Based Quantitative Trading Strategy

Project Objectives

Build a machine learning-based quantitative trading system that achieves stable excess returns in the S&P500 stock pool through multi-factor models and ranking learning algorithms.

Core Features

  • 160+ quantitative factors（Alpha101 + TA-Lib + Custom）
  • LightGBM Ranker ranking learning model
  • TopK Dropout portfolio optimization strategy
  • real-time trade execution（IBKR API）

Strategy Performance

Annual Return

-

Sharpe Ratio

-

Max Drawdown

-

Position Count

20

Technical Architecture

Data Layer
  • yfinance Data Fetching
  • DuckDB Data Storage
  • Parquet Format Storage
Model Layer
  • Factor Engineering (160+ factors)
  • LightGBM Ranker
  • Model Training & Validation
Trading Layer
  • Portfolio Optimization
  • IBKR Real-time Trading
  • Risk Control