Implementing a stock-exchange environment in OPEN-AI's gym environment.
To get started, you’ll need to have Python 3.5+ installed.
Simply install gym using pip:
pip install gym
Once gym is installed, clone this repository, then run
python3 demo_exchange.py
There are two versions, discrete and continuous action space -
agents may require one or the other - DDPG vs naive DQN, for example.
env = gym.make('game-stock-exchange-v0') is for a discrete environment.
env = gym.make('game-stock-exchange-continuous-v0') is for a continuous environment.
Code for continuous and discrete are found here.
If you want to integrate reinforcement learning agents, I recommend using
baselines, or stable baselines.
These are based on tensorflow.
If you're a fan of pytorch, stock-exchange-pytorch supports
implementations for reinforcement learning and supervised learning. This repository is actually an offshoot
of that project.
With a random agent, an output of demo_stock_exchange.py is like so:
With stable-baselines, Advantage Actor Critic (A2C) is used in the demo_exchange.py to train the model.
An output of such is like so:
