Knowledge Maxxing← Back to topic library
ML MaxxingBeginner60 sec
Feature
A feature is one measurable attribute of a training example that a model uses as input to make its prediction.
Prerequisite: Training example
What matters
Choosing which features to include, and how to represent them, has a bigger effect on model performance than most algorithm choices, since the model can only ever use the information it is given.
Example
For predicting house prices, square footage, number of bedrooms and neighborhood are each separate features describing the same house.
Common misconception
A feature is not automatically useful just because it is easy to measure; irrelevant features can add noise rather than signal.
Recall question
Why can adding an irrelevant feature hurt a model instead of helping it?
Keywords and related concepts
featureinput variableattributelabel