Local Outlier Factor#
Local Outlier Factor (LOF) measures the local deviation of density of a given sample with respect to its k nearest neighbors. As such, LOF only considers the local region (or neighborhood) of an unknown sample which enables it to detect anomalies within individual clusters of data.
Data Type Compatibility: Continuous
|1||k||20||int||The k nearest neighbors that form a local region.|
|2||contamination||null||float||The proportion of outliers that are presumed to be present in the training set.|
|3||tree||KDTree||object||The spatial tree used to run nearest neighbor searches.|
Return the base spatial tree instance:
public tree() : Spatial
use Rubix\ML\AnomalyDetection\LocalOutlierFactor; use Rubix\ML\Graph\Trees\BallTree; use Rubix\ML\Kernels\Distance\Euclidean; $estimator = new LocalOutlierFactor(20, 0.1, new BallTree(30, new Euclidean));
- M. M. Breunig et al. (2000). LOF: Identifying Density-Based Local Outliers.