Why Automated Feature Engineering Will Change the Way You Do Machine Learning


I've covered automated hyperparameter tuning a couple of times recently, as it's gotten a lot of attention with Google's AutoML. Automated feature engineering is a bit less commonly discussed. One of the authors of a new Python library, Featuretools, wrote this extensive post on what automated feature engineering is, why it works, and how to implement it using Featuretools.

The post is good, but if you really want to evaluate Featuretools for your own usage (which you should!), I'd dive in to the Jupyter notebooks here. They're very detailed, and get into exactly what features get created and how.

Automated feature engineering is something I personally want to play around with. I'm just a bit skeptical of the ability of an algorithm that has no contextual knowledge to do what has traditionally been a human task, but that's just as likely an indicator of my own egocentrism.


Want to receive more content like this in your inbox?