Trading Using Q-Learning

In this project, I will present an adaptive learning model to trade a single stock under the reinforcement learning framework. This area of machine learning consists in training an agent by reward and punishment without needing to specify the expected action. The agent learns from its experience and develops a strategy that maximizes its profits. This is my capstone project for the Machine Learning Engineer Nanodegree, from Udacity. You can check my report here and the notebook with the tests of the codes used in this project here. The TEX file was produced with help of Overleaf.


This project requires Python 2.7 and the following Python libraries installed:


In a terminal or command window, navigate to the top-level project directory QLearning_Trading/ (that contains this README) and run one of the following commands:

python qtrader/ <OPTION>
python -m qtrader.agent <OPTION>

Where OPTION could be train_learner, test_learner, test_random, optimize_k or optimize_gamma. The simulation will generate log files to be analyzed later on. Be aware that any of those commands take several minutes to finish.


  1. T.M. Mitchell. Machine Learning. McGraw-Hill International Editions, 1997. link
  2. M. Mohri, A. Rostamizadeh, A. Talwalkar. Foundations of Machine Learning. 2012. link
  3. N.T. Chan, C.R. Shelton. An Electronic Market-Maker. 2001 link
  4. N.T. Chan. Artificial Markets and Intelligent Agents. 2001 link
  5. R. Cont, k. Arseniy, and S. Sasha. The price impact of order book events. Journal of financial econometrics, 2014 link
  6. Du, Xin, Jinjian Zhai, and Koupin Lv. Algorithm Trading using Q-Learning and Recurrent Reinforcement Learning. link