ExtraTreesClassifier

Algorithm overview

Extremely randomized trees combine many trees whose split thresholds are sampled randomly. The added randomization can reduce variance while retaining nonlinear decision boundaries.

JavaScript implementation

Ensemble.ExtraTreesClassifier trains independent ExtraTreeClassifier members, averages their leaf class probabilities, and predicts the highest-probability class. It supports deterministic seeds and exposes averaged feature importances.

Quick start example

import { Ensemble } from '@kanaries/ml';
const model = new Ensemble.ExtraTreesClassifier({ nEstimators: 100, max_features: 'sqrt', randomState: 42 });
model.fit([[0, 0], [0, 1], [3, 3], [4, 3]], [0, 0, 1, 1]);
console.log(model.predict([[.2, .1], [3.5, 3]]), model.featureImportances);

Detailed API reference

Options include nEstimators?, bootstrap?, randomState?, max_features? (integer, fraction, sqrt, log2, or all), and the ExtraTreeClassifier depth/split options. Methods are fit(X, y), predictProba(X), and predict(X); classification averages leaf probabilities before taking the highest-probability class. classes and featureImportances are available after fitting.