Feature Engineering is the process of selecting, transforming, and creating the input features that a machine learning model uses, in order to improve its performance. It is often one of the most impactful parts of building a model.
What it means in plain English
Raw data rarely comes in the ideal form for a model to learn from. Feature engineering is the craft of shaping it — combining columns, creating new ones, scaling values, or extracting useful signals — so the patterns become easier for the model to detect. A well-engineered set of features can make a simple model outperform a complex one working with raw data.
For traditional machine learning it is a major, hands-on task; deep learning automates much of it, learning useful features directly from raw data.
A simple example
From a raw timestamp, you might engineer features like “day of week” or “is it a holiday,” which could reveal shopping patterns a model would miss if it only saw the raw date.
Why it matters
Feature engineering is where much of the human expertise in traditional machine learning lives. Good features often matter more than the choice of algorithm, which is why practitioners spend so much time on it.
Related terms
- Feature — the individual inputs feature engineering shapes.
- Dataset — the raw material for feature engineering.
- Deep Learning — automates much of feature engineering.
Frequently asked questions
What is feature engineering?
It is the process of creating, transforming, and selecting the input variables (features) that a model learns from, using domain knowledge to make patterns easier for the model to find.
Is feature engineering still needed with deep learning?
Deep learning can learn features automatically from raw data, reducing the need, but thoughtful feature engineering still helps in many practical and tabular-data problems.
