README.md

• GOAL:Alpha Generation
  • Research Problem：Multi-factor model based on machine learning
    • EDA of A share
      • benchmark :CSI300
    • Getting data from A share(Daily frequancy data)
    • Stock Data pre-procession
  • 1.Mutifactor Model
    • 1.1 Feature engineering(Factor mining)
      • Replicate factor from research report
      • some useful link
      • Research report
      • Interesting Paper (Trying use)
    • 1.2 Data preprocessing
    • 1.3 Single factor testing
    • 1.4 Multi-factor testing
    • 1.5 collinearity diagnostics
    • 1.6 Factor recombination
    • 1.7 Select stock from A share according to our model
      • Based on Machine learning
  • 2. Bulid Risk Model
  • 3. Portfolio optimization
  • 4.Build a Backtesting structure
  • TODO:
    • 1. Getting data and data preprocession
    • 2. Build a Backtesting structure and testing simple strategy
    • 3. Feature engineering
    • 4. Factor testing
    • 5. Select stock based on ML
    • 6. Portfolio optimization

GOAL:Alpha Generation

Research Problem：Multi-factor model based on machine learning

EDA of A share

benchmark :CSI300

Getting data from A share(Daily frequancy data)

• From Magnum Research

• From JQData

• From WIND(real-time data?)

Stock Data pre-procession

• Excluding ST shares, science and technology board, B shares, new shares, delisted shares.
• Exclude data with less than 14 days of trading per month.
• Remove missing values.
• Indent the data by 1% up or down.

1.Mutifactor Model

1.1 Feature engineering(Factor mining)

Replicate factor from research report

some useful link

1. GTJA 191 【研报复现】国泰君安——191个短周期量价特征因子选股系列一 - 知乎 https://zhuanlan.zhihu.com/p/30195354 日频量化因子分类 - 知乎 https://zhuanlan.zhihu.com/p/270265896

2. Alpha 101 数值型因子的大规模分层测试---WorldQuant 101、国泰191、Sundays100+ - 知乎 https://zhuanlan.zhihu.com/p/60872286

Research report

https://github.com/Barca0412/Introduction-to-Quantitative-Finance?tab=readme-ov-file

https://github.com/hugo2046/QuantsPlaybook

Interesting Paper (Trying use)

1.Generating Synergistic Formulaic Alpha Collections via Reinforcement Learning

https://arxiv.org/abs/2306.12964

2.AutoAlpha: an Efficient Hierarchical Evolutionary Algorithm for Mining Alpha Factors in Quantitative Investment https://arxiv.org/abs/2002.08245

1.2 Data preprocessing

• Depolarizing data
• Standardization data
• Neutrelizing data

1.3 Single factor testing

• Effectiveness(t-value&IC)
• Monotone Stability:Group backtest method

1.4 Multi-factor testing

1.5 collinearity diagnostics

1.6 Factor recombination

• single regression weighting
• IC,RankIC weighting
• Ranknet

1.7 Select stock from A share according to our model

Based on Machine learning

• Linear Model
• svm
• Lasso regression
• Random Forest
• Xgboost
• Catboost

2. Bulid Risk Model

3. Portfolio optimization

MAFS 5310 https://bookdown.org/shenjian0824/portr/#structure-of-the-book

• Mean-variance portfolio
• Markowitz portfolio
• Maximum Sharpe ratio portfolio
• Risk based portfolio ...

4.Build a Backtesting structure

Useful link for build a Backtesting structure

• backtrader
• backtrader中文教程
• Qlib
• QUANTAXIS 2.0.0
• QuantLib: the free/open-source library for quantitative finance

TODO:

1. Getting data and data preprocession

• get_data.py
• preprocess.py
• data: To store data

2. Build a Backtesting structure and testing simple strategy

• backtesting.py
• simple_backtest.ipynb；To test backtest structure

3. Feature engineering

• GTJA_191.py
• Alpha_101.py
• Myself_factor.py
• factor_data To store all factor data

4. Factor testing

• factor_analysis.py

5. Select stock based on ML

• ML_selector.py

6. Portfolio optimization

• portfolio_opt.py