Skip to main content

CS7646-ML4T

Machine Learning for Trading​

CS7646 ML4T

Book References​

machine-learning-for-trading

What Hedge Funds Really Do by Romero and Balch

Machine Learning, Tom Mitchell

Random Forest & Q-Learner Strategy Learner​

Final project for 3 different types of ML: Decision Trees, reinforcement learning, optimization.

Quantative factors X, dependant variable Y.

Regression Trees​

X1, X2, X3 -> Regression model -> Y

  1. Roll back to the beginning of data to get training data.
  2. Measure X1, X2, X3 data, not peek forward, only use data until the date.
  3. Build the model: Decision Trees, reinforcement learning, optimization.

Drawbacks:

Cannot distinguish between

  • Uncertain large return
  • Certain small return
  • When doing nothing is smart

Don't know when to exit.

Reinforcement learning​

Q&A​

What should we consider in manual strategy?​

Align the API.

Is it useful normalizing the indicators?​

Decision Tree could have not standardized indicators.

KNN needs to have standardized indicators.

How to make Dyna run faster?​

Use vectorizing code.

How do we combine indicators, sequentially?​

Or how do we assign weight or priority in developing a strategy?

  • Find threshold for each indicators. Just use the threshold, then voting
  • Sequentially one by one (All should be true)
  • Assign weight to each one.

Start with 1 indicator first, see if it's better to add more indicators.

Extra Resources​

Financial, Investing, and Trading​

Machine Learning​

  • Machine Learning (Book) - Tom Mitchell's concise introduction to machine learning, including supervised and reinforcement learning. Supplemental reading in CS7646, but required reading in CS7641 (OMSCS ML course)
  • An Introduction to Statistical Learning (Book) - A very good introductory ML book. Many "suggested" readings shared on Ed Discussions were from this book. (The second edition - hard cover - is expected to be released this summer, with a free PDF available in early 2022.)
  • Elements of Statistical Learning (Book) - A more mathematical and complete introduction to ML. This book may be used in one of the courses in the OMSA program.
  • Foundations of Machine Learning (Book) - Good material, but one of the more theoretical treatments of ML.
  • Probabilistic Machine Learning: An Introduction (Book) - The long-anticipated second edition of Kevin Murphy's Machine Learning Book. Will be a two-book series. Book 1 is camera-ready and book 2 will be released as a draft this summer. In some places, Murphy's book is used as an alternative to Tom Mitchell's book.
  • CS7641 Machine Learning (Georgia Tech OMSCS Course Videos)
  • CMU-601 Machine Learning (Carnegie Mellon University Course Videos)
  • Introduction to Machine Learning (University of Waterloo Course Videos)

Reinforcement Learning​

Deep Learning​

  • Deep Learning (Book) - This is the current leading Deep Learning textbook
  • Deep Learning Revolution (Book) - An easy-to-read book that presents Deep Learning, starting with a historical perspective and stepping forward to today.
  • Deep Learning (Nature Article) - A good article by Hinton and others on deep learning. Access through the Georgia Tech Library.
  • Grokking Deep Learning (Book) - A good book that, like ML4T, will have you implementing algorithms and models.
  • Foundations of Deep Reinforcement Learning (Book) - Sections 1.1 through 1.4 provide a nice introduction to reinforcement learning and our work in CS7646.
  • deeplearning.ai (Website) - A learning & education site by Andrew Ng.

PORTFOLIO VISUALIZER/efficient frontier