Glossary term

K-nearest neighbors

A method that predicts based on the labels of the closest examples in the data.

How it works

To classify a new point, k-nearest neighbors finds the k most similar training points and takes a majority vote, or averages their values for regression. There is no training step; it just stores the data. It depends on a good notion of distance.

A new point takes the label of its nearest neighbors

In code

K-nearest neighbors

from sklearn.neighbors import KNeighborsClassifier

knn = KNeighborsClassifier(n_neighbors=5).fit(X_train, y_train)

Why it matters

It is one of the simplest ideas in machine learning and a useful baseline.

Related terms

Classification Vector search Embedding