factor-mining

What problem does it solve?

This Skill provides an end-to-end platform for discovering, evaluating, and admitting alpha factors across multiple markets by combining large language models with reinforcement learning, reducing manual factor engineering work.

Core Features & Use Cases

• Automates generation of candidate factor formulas from Alpha101/Alpha191 seeds using LLM prompts.
• Evaluates candidates via GPU-accelerated IC/ICIR analysis, backtesting, and correlation checks, and manages admissions into a factor repository.
• Supports multi-market data pipelines (A股 astock and crypto markets) with market-specific data loading, calibration, and reporting workflows.

Quick Start

1. Prepare your environment by installing dependencies and running the provided scripts in scripts/ (e.g., batch_mine_parallel.py) to generate a batch of candidate factors.
2. Run a sample evaluation using the recommended market (e.g., astock) and inspect the admissions in the repository.
3. Use the professional_report.py tool to generate a factor analysis tear sheet for top factors.