Reinforcement Learning for trading securities

Note: Please Document any changes before requesting a commit to the master repo

For more details contact: ishan.khanka@gmail.com

Dir Structure

• ./config.ini : contains configuration
• ./data / : Location of input files (change input location in config)
• ./env/securities_trading_env.py : Env for agent
• ./modules/preprocessing.py : Making concat.csv
• ./main.py : Driver code
• ./save/ : Location for saved models

Flag Information

• "-l", "--load" default = "no_path" --> Only load the model
• "-v", "--verbose" default = 1 --> Flag for verbose either 1 or 0

To-Do:

• general-
  • reward = net worth ?
  • add a feature to run multiple bots
• env-
  • Add Connection to RabbitMQ

Fixed/ Done

• general-
  • Moved policy to config file
  • Added TF3
  • Added option to start agent with a specified observation window in config.ini
  • Added config.ini to reduce the number of flags
  • Added Hold Option
  • Create basic GYM environmet for trading securities
  • Added stable baselines for RL
  • Added flags for model selction
  • Added ability to save models with unique key based on time and date

Parent GitHub Link for archives and version control: https://github.com/gskishan004/Reinforcement-Learning-for-trading-securities

Common errors

• MPI : https://stackoverflow.com/questions/14004457/error-loading-mpi-dll-in-mpi4py