An example of Reinforcement Trading using OpenAI Baseline.
This is made by Adrian Portabales based on Mr. Peter Henry and inspired by Tito Ingargiola, hackthemarket.
There are some improvements in code:
-
4 States:
- 0 = Holding the trade. Only recalculate reward
- 1 = Buy
- 2 = Sell
- 3 = Close position
The bot only operates with one open position at the same time.
-
At the end of every episode, a test is run. The size of train and test set in
train_split
variable -
Now you can add more features (like indicators, momentums etc) with feature_engineering module.
OpenAI Baseline: https://github.com/openai/baselines
Gym_trading by Peter Henry: https://github.com/Henry-bee/gym_trading/
Also, good thougths in Trading and Q learning: https://github.com/savourylie/Stock-Price-Forecaster
Thanks to all!
Requirements: -Pandas -Matplotlib -Numpy -gym -TA-Lib -baselines
This framework was built in Python 3.5.2
Twitter: @porta4k
Mail: [email protected]