This projects aims to provide a Reinforcement-Learning (RL) approach to the Tactical Asset Allocation (TAA) problem. Given a financial portfolio of multiple assets, the optimal and dynamic allocation of the assets is a common problem in active portfoilo management. TAA involves the identification of market inefficiencies to exploit market timing effects by short term reallocation of the asset weights.
Table of Contents
To change the configurations of the environment you can either change them directly using the config.json file or run the config.py file using the flags listed below. By calling the config.py file a new config file mod_config.json will be created. If you want to go back to the default setting just delete that config file. Changes to the data basis can also been made through the config.json file but it is not recommend. Otherwise you have to train the supportive prediction models again.
Data can be requested using pandas_datareader. So far this project supports Yahoo Finance, Federal Reserve of Economic Data (FRED), The Investors Exchange, Moscow Exchange and Stooq as data resources. Only the (adjusted if available) closing prices will be tracked.
Note
Unfortunately there are often changes in the APIs of those website so I can't guarantee that those sites are still being supported some time onwards. At the time writing those supported websites are indeed working as intended.
Warning
Currently TensorFlow only supports Python version <=3.6.
- Free software: MIT License
- Documentation: https://pgtaa.readthedocs.io.
- TODO
Please take a look at Contributing.
This package was created with Cookiecutter and the audreyr/cookiecutter-pypackage project template.