Art of Feature Engineering

  title={The Art of Feature Engineering: Essentials for Machine Learning},
  author={Duboue, Pablo},
  publisher={Cambridge University Press},

Book Author

In my work as part of the IBM Watson Jeopardy! team, I created a custom programming language for feature engineering and witnessed how the impact of feature engineering outperformed any changes on the underlining models tried by machine learning colleagues.

Since then, I have been fascinated by feature engineering. I have seen many practitioners leave substantive improvements in model performance by wrongly focusing on finding better models rather than improving their features.

This book is the result of some comprehensive background research in the topic. It references more than 300 sources and it is targeted towards practitioners in the field looking to improve their feature engineering game.

To cater different learning styles, the book also includes 5 chapters of case studies, discussing common feature engineer techniques in fields such as graph data, text and images.

See the book companion's website for details.