Struct IsolationForest 

#[non_exhaustive]
pub struct IsolationForest { /* private fields */ }

Isolation Forest for anomaly detection.

Builds an ensemble of random isolation trees. Points that are isolated in fewer splits (shorter path lengths) receive higher anomaly scores.

Algorithm

1. Fit: Build n_estimators isolation trees, each trained on a random subsample of max_samples points.
2. Score: For each point, compute the average path length across all trees. Normalize to [0, 1] using score = 2^(-E[h(x)] / c(max_samples)).
3. Label: Scores above the threshold (determined by contamination) are labelled as anomalies (-1), others as normal (1).

Examples

use scry_learn::anomaly::IsolationForest;

let mut ifo = IsolationForest::new()
    .n_estimators(100)
    .contamination(0.05);
ifo.fit(&data).unwrap();
let labels = ifo.predict_labels(&data);

Implementations

impl IsolationForest

pub fn new() -> Self

Create a new IsolationForest with default parameters.

Defaults: 100 estimators, 256 max samples, 0.1 contamination.

pub fn n_estimators(self, n: usize) -> Self

Set the number of isolation trees (default: 100).

pub fn max_samples(self, n: usize) -> Self

Set the subsample size per tree (default: 256).

pub fn contamination(self, c: f64) -> Self

Set the expected contamination ratio (default: 0.1).

Must be in (0, 0.5]. Used to determine the anomaly score threshold.

pub fn random_state(self, seed: u64) -> Self

Set the random seed for reproducibility.

pub fn seed(self, s: u64) -> Self

Alias for random_state (consistent with other models).

pub fn fit(&mut self, features: &[Vec<f64>]) -> Result<()>

Build isolation trees on the provided feature matrix.

features is row-major: features[sample_idx][feature_idx].

Errors

Returns ScryLearnError::EmptyDataset if features is empty. Returns ScryLearnError::InvalidParameter if contamination is out of range.

pub fn predict(&self, features: &[Vec<f64>]) -> Vec<f64>

Compute anomaly scores for each sample.

Returns a Vec<f64> where values closer to 1.0 indicate anomalies and values closer to 0.0 indicate normal points.

Score formula: score = 2^(-E[h(x)] / c(ψ)) where E[h(x)] is the average path length and c(ψ) is the average BST path length for the training subsample size ψ.

pub fn predict_labels(&self, features: &[Vec<f64>]) -> Vec<i8>

Predict anomaly labels: 1 for normal, -1 for anomaly.

Uses the contamination-based threshold computed during fit. Returns all 1 (normal) if the model has not been fitted.

pub fn score_threshold(&self) -> f64

Returns the anomaly score threshold determined during fit.

Trait Implementations

impl Clone for IsolationForest

fn clone(&self) -> IsolationForest

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for IsolationForest

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for IsolationForest

fn default() -> Self

Returns the “default value” for a type.

impl Tunable for IsolationForest

fn set_param(&mut self, name: &str, value: ParamValue) -> Result<()>

Apply a named hyperparameter.

fn clone_box(&self) -> Box<dyn Tunable>

Clone this model into a boxed trait object.

fn fit(&mut self, data: &Dataset) -> Result<()>

Train on a dataset.

fn predict(&self, features: &[Vec<f64>]) -> Result<Vec<f64>>

Predict on row-major features.