Skip to main content

​
Overview

LogisticRegression implements logistic regression using gradient descent or L-BFGS optimization with native Zig acceleration. It supports both binary and multiclass classification via a one-vs-rest strategy.

​
Constructor

import { LogisticRegression } from "bun-scikit";

const model = new LogisticRegression(options);

​
Parameters

​
options
LogisticRegressionOptions
default:"{}"
Configuration options for the model

​
Methods

​
fit

Fit the logistic regression model using training data. 
fit(X: Matrix, y: Vector, sampleWeight?: Vector): this
Parameters:
  • X: Training data matrix of shape [nSamples, nFeatures]
  • y: Target class labels vector of length nSamples
  • sampleWeight: Optional sample weights (not currently used)
Returns: The fitted model instance Example: 
const X = [[0], [1], [2], [3], [4], [5]];
const y = [0, 0, 0, 1, 1, 1];

const model = new LogisticRegression({
  learningRate: 0.2,
  maxIter: 40_000,
  tolerance: 1e-10,
});
model.fit(X, y);

​
predict

Predict class labels for samples. 
predict(X: Matrix): Vector
Parameters:
  • X: Samples matrix of shape [nSamples, nFeatures]
Returns: Predicted class labels vector of length nSamples Example: 
const predictions = model.predict([[-1], [6]]);
console.log(predictions); // [0, 1]

​
predictProba

Predict class probabilities for samples. 
predictProba(X: Matrix): Matrix
Parameters:
  • X: Samples matrix of shape [nSamples, nFeatures]
Returns: Probability matrix of shape [nSamples, nClasses] Example: 
const proba = model.predictProba([[2.5]]);
console.log(proba[0]); // [0.4-0.6, 0.4-0.6] for boundary case

​
score

Return the mean accuracy on the test data. 
score(X: Matrix, y: Vector): number
Parameters:
  • X: Test samples matrix
  • y: True class labels
Returns: Mean accuracy (0 to 1)

​
Attributes

After fitting, the following attributes are available:
​
coef_
Vector | Matrix
Coefficient array:
  • For binary classification: Vector of length nFeatures
  • For multiclass: Matrix of shape [nClasses, nFeatures]
​
intercept_
number | Vector
Intercept (bias) term:
  • For binary classification: single number
  • For multiclass: Vector of length nClasses
​
classes_
Vector
Unique class labels encountered during fit
​
fitBackend_
'zig'
The backend used for fitting (always 'zig')
​
fitBackendLibrary_
string | null
Path to the native library used for fitting

​
Complete Example

​
Binary Classification

import { LogisticRegression } from "bun-scikit";

const X = [[0], [1], [2], [3], [4], [5]];
const y = [0, 0, 0, 1, 1, 1];

const model = new LogisticRegression({
  learningRate: 0.2,
  maxIter: 40_000,
  tolerance: 1e-10,
});

model.fit(X, y);

// Predictions
const preds = model.predict(X);
const accuracy = preds.filter((pred, idx) => pred === y[idx]).length / y.length;
console.log("Accuracy:", accuracy); // >0.99

// Probabilities
const proba = model.predictProba([[2.5]])[0][1];
console.log("P(class=1) at x=2.5:", proba); // ~0.5 (decision boundary)

​
Multiclass Classification

import { LogisticRegression } from "bun-scikit";

const X = [
  [0, 0],
  [0.1, 0.2],
  [0.2, -0.1],
  [2.0, 2.1],
  [2.2, 1.9],
  [1.8, 2.0],
  [4.0, 4.2],
  [3.9, 4.1],
  [4.2, 3.8],
];
const y = [0, 0, 0, 1, 1, 1, 2, 2, 2];

const model = new LogisticRegression({
  solver: "lbfgs",
  maxIter: 60,
  tolerance: 1e-6,
  l2: 0.01,
});

model.fit(X, y);
console.log("Score:", model.score(X, y)); // >0.95

const proba = model.predictProba([[0.1, 0.1], [2.1, 2.0], [4.1, 4.0]]);
console.log("Probabilities shape:", proba[0].length); // 3 classes

​
Notes

  • LogisticRegression requires native Zig kernels. Build them with bun run native:build if not using prebuilt binaries.
  • For multiclass problems, the model uses a one-vs-rest (OvR) strategy.
  • The L-BFGS solver (solver='lbfgs') typically converges faster than gradient descent but requires model-handle native kernels.
  • L2 regularization helps prevent overfitting, especially with high-dimensional data.

Build docs developers (and LLMs) love

Get started for freeTalk to us