k-Nearest Neighbors

noun

say it: “KAY NEER-ist NAY-berz” /ˈkeɪ ˈnɪrəst ˈneɪbərz/

A method that classifies a new example by finding the k most similar examples it already knows and taking their most common answer.

What is k-nearest neighbors?

k-nearest neighbors (k-NN) guesses by finding the k training examples closest to a new one, by distance, and taking their most common label, or their average for a number. It learns nothing ahead of time: it keeps the examples and compares when it guesses. scikit-learn's is KNeighborsClassifier.

from sklearn.neighbors import KNeighborsClassifier
fruit = [[120, 3], [125, 4], [130, 3], [180, 8], [185, 9], [190, 8]]   # grams, sweetness
names = ["lemon", "lemon", "lemon", "orange", "orange", "orange"]
knn = KNeighborsClassifier(n_neighbors=3).fit(fruit, names)
print(knn.predict([[128, 4]]))   # ['lemon']: its 3 nearest are lemons

In a sentence

  • “k-Nearest neighbors needs scaled features, because it decides by distance.”
  • “With k set to 5, k-nearest neighbors looked at the five most similar fruits and took a vote.”

Level: Journeyman · More AI Basics terms at this level

Updated