Machine Learning for Trading (wip)

Description

Installation

Requires python (preferably >=3.11) and git to be installed.

git clone https://github.com/Ikeafisch3/ML4T.git
pip install -r requirements.txt

For CUDA GPUs use pip install -r requirements-gpu.txt

Installing IBKR Gateway

Refer to IBKR webpage.

or

If Java is installed, download the zip file from here and extract it to .\ibkr\gateway. All files should be in \ML4T\ibkr\gateway\clientportal.gw\.

Start a terminal in .ibkr\gateway\clientportal.gw\ and run bin\run.bat root\conf.yaml. Login with your Paper Trading account at https://localhost:[port] (by default the URL is https://localhost:5000, however this can be changed in the root\config.yaml)

After this ibkr\api.py should be functional.

TODO

• Determine a more fitting model for the task
• Add Paper Trading capability
• Add designated training scripts
• Add the following indicators:

  • Relative Strength index (RSI) --> speed and change of price movements

  • Stochastic Oscillator (compares closing price with severel key historical prices (highs, lows)) --> indicates momentum

  • put / call ratio --> not so relevant for large cap but might include anyway

  • some general economic indicator related to gdp growth, unemployment rate etc. --> use "Public Data API" by the U.S. BUREAU OF LABOR STATISTICS

  • Index Performances: SPX, sector indexes, emerging markets, small cap indexes

  • bid/ask spread --> provides insight into market liquidity and volatility

  • Days until next financial data release or annual shareholder meeting

  • previous candles

• Fixes:
  • properly implement fear and greed data (eg. for different candle sizes)
  • rate limits on IBKR Gateway are not properly implemented (should be different for some requests)
  • inference.py needs clean up and better docs