Note that you can add new factor calculation methods and construct your own factor pool, then play with the combination tools and the evaluation tools to renew your factor or to verify its effect. Every tool in this project is open to rebuild, just add your innovative idea and creativity!

This project depends on:

• akshare

• matplotlib

• numpy

• pandas

• seaborn

• statsmodels

• scipy

• nri_quant

• torch

There are two jupyter notebooks. single_factor.ipynb helps you to analyze a single factor, and multi_factors.ipynb integrates tools to load a calculated factor pool and factors selection, combination and evaluation.

Our evaluation method includes:

• distribution plot
• check if distribution is close to normal
• adf test
• ic test
• Grangers causation test
• trading back test
• layered back test

• factors pool loading:

  • Calculated factors are saved as csv files located in the path ./factorloader/data

• after factor combination:

  • Combination weights are located in the path ./factorcombiner/data

• ./stockdownload includes downloading stocks' original trading data and its return calculation for the factors combination tool.
• ./factorloader generates factors pool, all of the factor calculation methods are saved at factorgens.py and an instruction also included.
• ./factorcombiner saves both factors selection and combination tools.
• ./evaluationtools saves all of the evaluation methods.
• ./utils includes practical tools and a data's preprocessing tool.

You can check the handbook.pdf located at ./Summary presentation

Further learning materials:

• 利用python对国内各大券商的金工研报进行复现
• 择时论文复现笔记