IsolationForest

Isolation Forest for outlier detection Compatible with sklearn.ensemble.IsolationForest

Detects outliers using ensemble of isolation trees. Outliers are isolated closer to the root of the tree.

Example

const iso = new IsolationForest({ contamination: 0.1, n_estimators: 100 });
iso.fit(X_train);
const predictions = iso.predict(X_test);  // -1 for outliers, 1 for inliers
const scores = iso.score_samples(X_test); // Anomaly scores

Constructors

Constructor

new IsolationForest(options?): IsolationForest

Defined in: ds/src/ml/outliers.js:252

Parameters

options?

contamination

number = 0.1

Expected proportion of outliers (default: 0.1)

label_column

string = "outlier"

Name of output column for predictions (default: ‘outlier’)

max_features

number = 1.0

Features to draw for each tree (default: 1.0 = all)

max_samples

number = "auto"

Samples to draw for each tree (default: ‘auto’ = min(256, n))

n_estimators

number = 100

Number of trees (default: 100)

random_state

number = null

Random seed (default: null)

Returns

IsolationForest

Properties

_groupModels

_groupModels: Map<any, any>

Defined in: ds/src/ml/outliers.js:274

_originalData

_originalData: any

Defined in: ds/src/ml/outliers.js:273

_tableColumns

_tableColumns: any

Defined in: ds/src/ml/outliers.js:272

contamination

contamination: number

Defined in: ds/src/ml/outliers.js:262

label_column

label_column: string

Defined in: ds/src/ml/outliers.js:265

max_features

max_features: number

Defined in: ds/src/ml/outliers.js:263

max_samples

max_samples: number

Defined in: ds/src/ml/outliers.js:261

max_samples_

max_samples_: number

Defined in: ds/src/ml/outliers.js:268

n_estimators

n_estimators: number

Defined in: ds/src/ml/outliers.js:260

nFeatures_

nFeatures_: number

Defined in: ds/src/ml/outliers.js:271

offset_

offset_: number

Defined in: ds/src/ml/outliers.js:269

random_state

random_state: number

Defined in: ds/src/ml/outliers.js:264

threshold_

threshold_: number

Defined in: ds/src/ml/outliers.js:270

trees_

trees_: any[]

Defined in: ds/src/ml/outliers.js:267

Methods

_fitSingleModel()

_fitSingleModel(X): any

Defined in: ds/src/ml/outliers.js:354

Internal method to fit a single isolation forest model

Parameters

X

number[][]

2D array of numeric data

Returns

any

Model parameters

_predictWithModel()

_predictWithModel(X, model): number[]

Defined in: ds/src/ml/outliers.js:591

Internal method to predict with a specific model

Parameters

X

number[][]

Data

model

any

Model parameters

Returns

number[]

Predictions

fit()

fit(X): IsolationForest

Defined in: ds/src/ml/outliers.js:282

Fit the model

Parameters

X

any

Training data (2D array or {data, columns, group})

Returns

IsolationForest

this

fit_predict()

fit_predict(X): any[] | number[]

Defined in: ds/src/ml/outliers.js:632

Fit and predict in one step (sklearn compatibility)

Parameters

X

any

Data (2D array or {data, columns, group})

Returns

any[] | number[]

Predictions: -1 for outliers, 1 for inliers (or table with outlier column)

fit_transform()

fit_transform(X): any[] | number[]

Defined in: ds/src/ml/outliers.js:623

Fit and transform in one step Primary API for outlier detection with tables

Parameters

X

any

Data (2D array or {data, columns, group})

Returns

any[] | number[]

Labels or table with outlier column

predict()

predict(X): any[] | number[]

Defined in: ds/src/ml/outliers.js:499

Predict if samples are outliers

Parameters

X

any

Data (2D array or {data, columns, group})

Returns

any[] | number[]

Predictions: -1 for outliers, 1 for inliers (or table with outlier column)

score_samples()

score_samples(X): number[]

Defined in: ds/src/ml/outliers.js:426

Compute anomaly scores for samples Lower (more negative) scores indicate outliers Scores range approximately from -1 to 0

Parameters

X

any

Data (2D array, {data, columns}, or array of objects)

Returns

number[]

Anomaly scores (negative values)

transform()

transform(X): any[] | number[]

Defined in: ds/src/ml/outliers.js:613

Transform data by adding outlier labels Alias for predict() - primary API for table-based workflows

Parameters

X

any

Data (2D array or {data, columns, group})

Returns

any[] | number[]

Labels or table with outlier column