Your project will be building a rebalanced portfolio(50 - 100 holdings) based on the prediction only on 13F data(https://www.investopedia.com/terms/f/form-13f.asp) and daily price data in Python(attached in this link https://drive.google.com/open?id=1zmz_EjO2GLFvAAvpSnIy3At3wOmyb37o). Rebalancing period (prediction period) is preferravblelly quarterly, but also can be monthly, weekly or daily. There are two SQL databases. Following links are reference on SQL:
- Construct sql databases by MYI MYD FRM files https://stackoverflow.com/questions/879176/how-to-recover-mysql-database-from-myd-myi-frm-files
- Install sql on Mac https://dev.mysql.com/doc/mysql-osx-excerpt/5.7/en/osx-installation-pkg.html
- Software to view sql data https://www.mysql.com/products/workbench/
- SQL query https://www.hackerrank.com/domains/sql
- Connecting sql with Python by pandas or MySQLDB http://mysql-python.sourceforge.net/MySQLdb.html https://gist.github.com/stefanthoss/364b2a99521d5bb76d51
https://quantpedia.com/strategies/alpha-cloning-following-13f-fillings/ https://marcosammon.com/2016/08/08/13f.html https://whalewisdom.com/whitepapers/backtesting
Use just SP500 Stocks, use only managers with fewer than 100 holdings, and more than 5 holdings XGBoost, which hedge funds would outperform the most Pick a portfolio 5-15 stocks most held by those hedge fund managers
- can break up by industry
Construct alpha factors on the hedge funds
Can also do reinforcement learning
- agent is trying to pick 13Fs that are the most outperforming
Create a deep learning/reinforcement learning system to predict economic upturns/downturns
- South America: Brazil, Argentina, Chile, Columbia