Knowledge Maxxing← Back to topic library
ML MaxxingIntermediate75 sec
Feature engineering
Feature engineering is creating new input features, or transforming existing ones, to better expose the patterns a model needs to learn.
Prerequisite: Feature
What matters
Well-designed features can make a simple model perform as well as a much more complex one, since the model can only find patterns that its input features make visible.
Example
Turning a raw timestamp into separate 'day of week' and 'hour of day' features can expose weekly and daily patterns a model could not otherwise easily learn.
Common misconception
Feature engineering is not obsolete because of deep learning; even models that learn features automatically from raw data often benefit from thoughtfully engineered inputs.
Recall question
Why can a well-engineered feature let a simple model match a more complex one?
Keywords and related concepts
feature engineeringfeature transformationpreprocessingfeature