Isolation Forest (JS / TS)

Isolation Forest for JavaScript / TypeScript.

Use it to efficiently detect anomalies in your dataset.

Install

$ npm install --save isolation-forest

Usage

import { IsolationForest } from 'isolation-forest'

var isolationForest = new IsolationForest();
isolationForest.fit(trainingData) // Type ObjectArray ({}[]); 

var trainingScores = isolationForest.scores()

// then predict any data
var scores = isolationForest.predict(data)

Parameters

IsolationForest(numberOfTrees, subsamplingSize) takes 2 optional paramaters.

numberOfTrees: Amount of trees the forest should generate. Default is 100, because for most datasets the anomaly scores converges with less than 100 trees.

subsamplingSize: Size of used subsamples of the dataset during trainging phase. Helps avoiding common problems in anomaly detection (swamping and masking). Default is 256 or the dataset size, if smaller.

For more info, see [1].

Anomaly score

As stated in [1]:

If instances return a score very close to 1, then they are definitely anomalies.

If instances have a score much smaller than 0.5, then they are quite safe to be regarded as normal instances.

If all the instances return a score ≈ 0.5, then the entire sample does not really have any distinct anomaly.

Misc

Thank you for using my package. If you find bugs or have ideas to improve the package, open a PR on GitHub with your changes. I'll add the changes as soon as possible.

References

1. Liu, F. T., Ting, K. M., & Zhou, Z. H. (2008, December). Isolation forest. In Data Mining, 2008. ICDM’08. Eighth IEEE International Conference on (pp. 413–422). IEEE.