RadiusNeighborsRegressor

Algorithm overview

RadiusNeighborsRegressor predicts continuous values from all training samples within a fixed radius. It can be more stable than K-neighbor regression when sample density varies across the feature space.

JavaScript implementation

@kanaries/ml provides Neighbors.RadiusNeighborsRegressor for JavaScript applications with uniform or distance-weighted averaging.

Quick start example

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

const reg = new Neighbors.RadiusNeighborsRegressor({ radius: 1.25, weights: 'distance' });
reg.fit([[0], [1], [4]], [0, 1, 4]);
const pred = reg.predict([[0.5], [10]]);

Detailed API reference

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

Methods:

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

If no neighbors are found for a query sample, the prediction is Number.NaN.