KNeighborsRegressor

Algorithm overview

KNeighborsRegressor predicts a continuous value by finding nearby training samples and averaging their targets. It is useful as a non-parametric baseline when similar examples should have similar outcomes.

JavaScript implementation

@kanaries/ml provides Neighbors.KNeighborsRegressor for browser and Node.js workflows. It supports uniform or distance weighting and the distance metrics exposed by Metrics.Distance.

Quick start example

import { Neighbors } from '@kanaries/ml';

const reg = new Neighbors.KNeighborsRegressor({ nNeighbors: 2, weights: 'distance' });
reg.fit([[0], [1], [3]], [0, 1, 3]);
const pred = reg.predict([[2]]);

Detailed API reference

new Neighbors.KNeighborsRegressor()
new Neighbors.KNeighborsRegressor(nNeighbors: number, weights?: 'uniform' | 'distance', metric?: Distance.IDistanceType, p?: number)
new Neighbors.KNeighborsRegressor(props?: {
  nNeighbors?: number;
  weights?: 'uniform' | 'distance';
  metric?: Distance.IDistanceType;
  p?: number;
})

Methods:

• fit(trainX: number[][], trainY: number[]): void
• predict(testX: number[][]): number[]

metric defaults to 'euclidean', nNeighbors defaults to 5, and p is used by Minkowski distance.