Overview
The KNeighborsRegressor implements k-nearest neighbors regression. It predicts the value of a sample based on the average (or weighted average) of the values of its k nearest neighbors in the training set.
Constructor
import { KNeighborsRegressor } from '@scikitjs/sklearn';
const regressor = new KNeighborsRegressor({
nNeighbors: 5,
weights: 'uniform'
});
Parameters
Number of neighbors to use for regression. Must be a positive integer.
weights
'uniform' | 'distance'
default:"uniform"
Weight function used in prediction:
'uniform': All neighbors are weighted equally'distance': Neighbors are weighted by the inverse of their distance (closer neighbors have more influence)
Methods
fit()
Fit the k-nearest neighbors regressor from the training dataset.
fit(X: Matrix, y: Vector, sampleWeight?: Vector): this
Training data matrix where each row is a sample and each column is a feature.
Target values (continuous values) for the training data.
Sample weights (currently not implemented but reserved for future use).
this - The fitted regressor instance.
Throws:
- Error if
nNeighbors exceeds the training set size
- Error if input validation fails
predict()
Predict the target values for the provided data.
predict(X: Matrix): Vector
Vector - Predicted values.
score()
Return the coefficient of determination (R²) of the prediction.
score(X: Matrix, y: Vector): number
True values for the test samples.
number - R² score (coefficient of determination).
Examples
Basic Regression
import { KNeighborsRegressor } from '@scikitjs/sklearn';
// Training data
const X = [
[0], [1], [2], [3], [4]
];
const y = [0, 1, 4, 9, 16]; // y = x²
// Create and train regressor
const knn = new KNeighborsRegressor({ nNeighbors: 3 });
knn.fit(X, y);
// Predict new values
const predictions = knn.predict([[1.5], [2.5]]);
console.log(predictions); // [2.33, 6.33] approximately
Distance-Weighted Regression
import { KNeighborsRegressor } from '@scikitjs/sklearn';
// Training data with non-linear relationship
const X = [
[1.0], [2.0], [3.0], [4.0], [5.0]
];
const y = [2.0, 4.5, 7.0, 10.0, 13.5];
// Use distance weighting for better interpolation
const knn = new KNeighborsRegressor({
nNeighbors: 3,
weights: 'distance'
});
knn.fit(X, y);
// Predict intermediate values
const prediction = knn.predict([[2.5]]);
console.log(prediction); // Closer neighbors have more influence
Multi-dimensional Regression
import { KNeighborsRegressor } from '@scikitjs/sklearn';
// Housing prices based on [size, bedrooms]
const X = [
[1200, 2],
[1400, 3],
[1600, 3],
[1800, 4],
[2000, 4]
];
const y = [200000, 250000, 280000, 320000, 360000];
const knn = new KNeighborsRegressor({ nNeighbors: 3 });
knn.fit(X, y);
// Predict price for new house
const price = knn.predict([[1500, 3]]);
console.log(`Predicted price: $${price[0]}`);
Model Evaluation
import { KNeighborsRegressor } from '@scikitjs/sklearn';
// Training data
const XTrain = [
[1], [2], [3], [4], [5], [6]
];
const yTrain = [2, 4, 6, 8, 10, 12];
// Test data
const XTest = [[2.5], [4.5]];
const yTest = [5, 9];
const knn = new KNeighborsRegressor({ nNeighbors: 2 });
knn.fit(XTrain, yTrain);
// Calculate R² score
const r2 = knn.score(XTest, yTest);
console.log(`R² score: ${r2}`);
import { KNeighborsRegressor } from '@scikitjs/sklearn';
const X = [[1], [2], [3], [4], [5]];
const y = [1, 4, 9, 16, 25]; // y = x²
// Uniform weighting
const uniformKNN = new KNeighborsRegressor({
nNeighbors: 3,
weights: 'uniform'
});
uniformKNN.fit(X, y);
// Distance weighting
const distanceKNN = new KNeighborsRegressor({
nNeighbors: 3,
weights: 'distance'
});
distanceKNN.fit(X, y);
const testPoint = [[2.5]];
console.log('Uniform:', uniformKNN.predict(testPoint));
console.log('Distance:', distanceKNN.predict(testPoint));
// Distance weighting gives closer neighbors more influence
Notes
- Uses Euclidean distance to find nearest neighbors
- Distance weighting uses
1 / distance with a minimum threshold to avoid division by zero
- If a test sample coincides exactly with a training sample (distance = 0), that training value is returned directly
- Stores the entire training dataset (instance-based learning)
- Prediction time scales linearly with training set size
- Consider feature scaling when features have different units or ranges