Isolation Forest in JavaScript with @kanaries/ml

Learn what Isolation Forest does, when to use it, and how to run IsolationForest in JavaScript or TypeScript with @kanaries/ml for browser and Node.js applications.

Algorithm overview

IsolationForest detects anomalies by isolating rare points with shorter path lengths in random partition trees.

This algorithm is especially useful when:

You have mostly normal behavior with relatively few outliers.
Labeling anomalies is expensive or unavailable.
You need near-real-time anomaly scoring in browser or Node.js.

@kanaries/ml provides Isolation Forest in JavaScript for anomaly detection features that need to run inside a product, a browser-based monitoring view, or a Node.js scoring service. That is useful when you want outlier detection near the application layer, for example in fraud heuristics, telemetry triage, or unusual-behavior alerts.

Keeping anomaly scoring in JS makes it easier to connect feature extraction, threshold logic, and downstream product actions without introducing a Python serving dependency.

Quick start

IsolationForest in Python vs JavaScript / TypeScript

If you searched for "IsolationForest in JavaScript" or "IsolationForest in TypeScript", this section maps the familiar scikit-learn call to the equivalent @kanaries/ml usage for browser and Node.js runtimes.

Python

scikit-learn

from sklearn.ensemble import IsolationForest

X = [[0, 0], [0.1, 0.2], [0.2, 0.1], [8, 8]]

clf = IsolationForest(n_estimators=50, contamination=0.25, random_state=0)
clf.fit(X)
pred = clf.predict(X)

JavaScript / TypeScript

@kanaries/ml

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

const X = [[0, 0], [0.1, 0.2], [0.2, 0.1], [8, 8]];

const clf = new Ensemble.IsolationForest(256, 50, 0.25);
clf.fit(X);
const pred = clf.predict(X);

Quick JavaScript example

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

const X = [[0, 0], [0.1, 0.2], [0.2, 0.1], [8, 8]];

const clf = new Ensemble.IsolationForest(256, 50, 0.25);
clf.fit(X);
const pred = clf.predict(X);

Detailed API reference

constructor(subsampling_size: number = 256, tree_num: number = 100, contamination: 'auto' | number = 'auto')

Parameters

name	type	default	description
subsampling_size	number	256	Number of samples used to build each tree
tree_num	number	100	Number of isolation trees in the forest
contamination	'auto' \| number	'auto'	Expected proportion of outliers

Algorithm

IsolationForest randomly splits features to isolate samples. Points that can be isolated with fewer splits are considered anomalies.

Methods

fit(samplesX: number[][]): void
predict(samplesX: number[][]): (0|1)[]

Implementation workflow

Train on representative mostly-normal historical samples.
Predict anomaly labels or scores on incoming events.
Tune contamination and decision thresholds using alert precision targets.

JavaScript deployment notes

Train on representative mostly-normal data so anomaly scores reflect real deviations.
Treat contamination and alert thresholds as product decisions, not just model defaults.
For real-time scoring, keep feature generation and model prediction in the same Node.js pipeline when possible.

Isolation Forest