/DeepTrader

DeepTrader: Deep Reinforcement Learning TradeBot using DQN architecture

Primary LanguagePython

DeepTrader: Deep Reinforcement Learning TradeBot

Uses a DQN Network to train an agent to buy/short/hold positions in the futures markets.

Versions of the agent:

  • trade_bot_cuda.py: Long Only, Fully Connected Layer networks
  • trade_bot_fut_cuda.py: Long/Short, Fully Connected Layer networks
  • trade_bot_lstm_5_cuda.py: Long/Short, LSTM with Stop Loss and Full Trade Profit Reward
  • trade_bot_lstm_6_cuda.py: Long/Short, LSTM

All data input files are located in the ./data folder. Command arguments:

  1. Input data file path
  2. Output data file path
  3. Mode: [train] to train the agent and save the model params. [test] to test the agent on the previously saved model params.

To run:

python .\trade_bot_lstm5_cuda.py .\data\RTY=F.csv rty_lstm5 train
python .\trade_bot_lstm5_cuda.py .\data\RTY=F.csv rty_lstm5 test