Classification (Machine Learning)

Classification is a type of supervised machine learning in which a model learns from labelled examples to predict a discrete category for new inputs. With two categories it is binary classification; with more it is multiclass; when an input can have several labels at once it is multilabel.

Why it matters

Many practical problems are classification: spam filtering, fraud flags, ticket routing, image recognition and sentiment analysis. How you measure success matters. Accuracy can be misleading when classes are imbalanced, so precision, recall and F1 score are commonly reported instead, and the cost of false positives versus false negatives should drive the threshold you pick.

Example

Fraud detection: 1,000 transactions, 10 fraudulent
A model that always says "legitimate" is 99% accurate
but catches 0 of 10 frauds (recall = 0%).

This is why accuracy alone can hide a useless model.

Common mistakes

• Judging imbalanced problems by accuracy alone.
• Leaking test data into training, which inflates results.
• Confusing it with regression, which predicts a continuous number, and with clustering, which has no labels.

Related

• Supervised learning
• Regression
• Decision tree
• Data labeling

This page is general information written by our editorial team and updated when documentation changes. Servers and providers differ, so test changes safely first. See our editorial policy and disclaimer.