Linear Regression

Algorithm overview

LinearRegression models continuous targets with an interpretable linear relationship between features and outputs.

This algorithm is especially useful when:

• You need a transparent baseline for numeric prediction tasks.
• Feature-target relationships are approximately linear after transformation.
• You want fast training and low-latency inference in JavaScript.

JavaScript implementation

@kanaries/ml gives JavaScript teams a straightforward linear regression implementation for numeric prediction tasks that need to stay in the application layer. This is useful for browser-side demos, pricing calculators, forecasting helpers, and Node.js APIs that want transparent coefficients instead of opaque black-box behavior.

Because the model is simple and explainable, it fits especially well in product contexts where engineers and stakeholders need to reason about why a prediction changed.

Interactive linear regression playground

Change the dataset and noise, then click the chart to add your own observation. The fitted line and holdout metrics update immediately using Linear.LinearRegression in your browser.

Quick start

from sklearn.linear_model import LinearRegression

X = [[0], [1], [2], [3]]
y = [1, 3, 5, 7]

reg = LinearRegression()
reg.fit(X, y)
pred = reg.predict([[4], [5]])

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

const X = [[0], [1], [2], [3]];
const y = [1, 3, 5, 7];

const reg = new Linear.LinearRegression();
reg.fit(X, y);
const pred = reg.predict([[4], [5]]);

Quick JavaScript example

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

const X = [[0], [1], [2], [3]];
const y = [1, 3, 5, 7];

const reg = new Linear.LinearRegression();
reg.fit(X, y);
const pred = reg.predict([[4], [5]]);
console.log(pred);

Detailed API reference

constructor()

This class implements ordinary least squares linear regression. It estimates coefficients for a linear model by minimizing the squared error between predicted and actual values.

Methods

• fit(X: number[][], Y: number[]): void
• predict(X: number[][]): number[]

fit

• X - Feature matrix of shape [nSamples, nFeatures].
• Y - Target values of length nSamples.

predict

• X - Feature matrix for which to compute predictions.

Implementation workflow

1. Prepare numeric features and split into train/validation sets.
2. Fit the model and inspect residual patterns for systematic errors.
3. Iterate on feature engineering when residuals show non-linear structure.

JavaScript deployment notes

• Use linear regression as a baseline before moving to more complex non-linear models.
• Inspect residuals to decide whether feature engineering or a different model family is needed.
• This estimator is a strong fit for small to medium tabular problems where interpretability matters.