Course contentsShow
← Machine Learning and Deep Learning
Lesson 235 of 3,5384. Classification with Logistic RegressionFree lesson

What is Classification?

Understanding classification as predicting discrete categories rather than continuous values, with examples across domains.

What is Classification?

What you'll learn: You'll discover how classification differs from regression by predicting categories instead of numbers.

From Numbers to Categories

So far, you've been working with regression — predicting continuous numerical values like house prices or temperatures. But what if you want to predict something that falls into distinct categories?

Classification is a supervised learning task where the goal is to predict which discrete category (or "class") a data point belongs to, rather than a continuous number.

Everyday Classification Examples

Classification problems surround us:

  • Email filtering: Is this message spam or not spam? (2 categories)
  • Medical diagnosis: Does this patient have disease A, B, C, or is healthy? (4 categories)
  • Image recognition: Is this photo a cat, dog, or bird? (3 categories)
  • Credit approval: Should we approve or deny this loan application? (2 categories)

Notice how each prediction is a label from a fixed set of possibilities, not a number on a continuous scale.

Classification vs Regression

The key difference:

  • Regression: Predict a quantity → "This house will sell for $347,250"
  • Classification: Predict a category → "This email is spam" or "This tumor is benign"

Think of it like sorting mail into bins versus weighing packages. Classification puts things into labeled boxes; regression measures them on a scale.

Binary and Multiclass

Classification problems come in flavors:

  • Binary classification: Two possible outcomes (yes/no, spam/ham, positive/negative)
  • Multiclass classification: Three or more categories (cat/dog/bird, disease types A-E)

Key Takeaway: Classification predicts discrete categories rather than continuous values, transforming real-world "which type?" questions into machine learning problems.