Home › Key concepts › Feature engineering

Key concept · Walk (intermediate) · the same on every platform

Feature engineering

Turn raw data into signals a model can use.

What it is

Features are the inputs to a model: 'orders in the last 30 days', 'average basket size', 'days since last login'. Good features, computed only from data known at prediction time, matter more than the choice of algorithm for most tabular problems.

Why it matters

Most of the gain in classic ML comes from features, not models.

An example

'Number of support calls in the last 14 days' turns out to be the strongest churn signal.

The common mistake

Computing features with future data by accident.

Where you meet it on each platform

The idea stays the same; only the product name changes.

JobAWSAzureGoogle CloudDatabricksOpen source
Feature storethe same feature values for training and for live predictionsSageMaker Feature StoreAzure ML feature storeVertex AI Feature StoreFeature engineering in Unity CatalogFeast

Related ideas

Where it shows up

Part of the Walk level on these paths: ML Engineering (features and data).