Introduction

Introduction

A chef can only cook as well as the ingredients allow. In machine learning, the features are the ingredients. Feature engineering means creating features that make the pattern easier for the model to see.

Often, the raw columns aren't in the most useful form. "Total rooms in a district" isn't very informative on its own — a district may simply be big. "Rooms per household" tells you much more about how large homes are. Good feature engineering frequently improves a model more than switching to a fancier algorithm, especially for tabular data.

This topic covers three common techniques:

1. combining attributes;
2. transforming skewed features;
3. bucketizing numbers into ranges.