Course contentsShow
← Machine Learning and Deep Learning
Lesson 420 of 3,53810. Feature EngineeringFree lesson

One-Hot Encoding Fundamentals

Creating binary columns for each category value, the standard approach for nominal variables.

One-Hot Encoding Fundamentals

What you'll learn: How to transform nominal categorical variables into a format machine learning models can understand by creating separate binary columns for each category.

The Core Problem

Machine learning models work with numbers, not category names. When you have nominal variables (categories with no inherent order, like "red," "blue," "green"), you can't simply assign arbitrary numbers (1, 2, 3) because that implies an ordering that doesn't exist. One-hot encoding solves this problem elegantly.

How One-Hot Encoding Works

One-hot encoding creates a new binary column for each unique category value. Each column acts as an on/off switch (1 or 0) indicating whether that category is present.

Analogy: Think of it like light switches in a room. If you have three colors (red, blue, green), you install three switches. For a red item, only the "red" switch is ON (1), while "blue" and "green" switches are OFF (0).

Example Transformation

Original data:

Color
-----
Red
Blue
Green
Red

After one-hot encoding:

Color_Red  Color_Blue  Color_Green
---------  ----------  -----------
    1          0           0
    0          1           0
    0          0           1
    1          0           0

Why This Standard Approach?

One-hot encoding treats all categories equally with no implicit ordering. Each category gets its own dimension in the feature space, preventing the model from incorrectly learning relationships based on arbitrary numeric assignments.

Key Takeaway: One-hot encoding converts nominal categorical variables into binary columns (one per category), creating a numerical representation that preserves the categorical nature without introducing false ordinal relationships.