PPO2-currency-trader

Build reinforcement learning agents to trade USD_CAD and USD_CHF using data from OANDA v20 API.

After training, you can visualize the agents' trading activities (using testing data) on Dash app.

Credits: https://github.com/notadamking/RLTrader for gym environment codes, tutorials, and idea of using RL in trading https://dash-gallery.plotly.host/Portal/ for Dash template

1. Install dependencies
   • Make sure you have libopenmpi-dev

     $ sudo apt install libopenmpi-dev

   • If using GPU

     $ pip install -r requirements-gpu.txt

   • If not using GPU

     $ pip install -r requirements-nogpu.txt

2. Create a config file in ~/.v20.conf using information from your OANDA account. For example,

   hostname: api-fxpractice.oanda.com
   streaming_hostname: stream-fxpractice.oanda.com
   port: 443
   ssl: true
   token: XXXXX
   username: XXXXX
   datetime_format: RFC3339
   accounts:
   - XXX-XXX-XXXXXXX-XXX
   active_account: XXX-XXX-XXXXXXX-XXX

3. Download data from OANDA

   $ python data.py

4. Train models using PPO2. For example,

   $ python model.py --symbol EUR_USD --n 10000

5. Visualize agent training

   $ python app.py