KI-Glossar
Das vollständige Wörterbuch der Künstlichen Intelligenz
Local Outlier Factor (LOF)
Anomaly detection algorithm that measures the local density deviation factor of a point relative to its neighbors. An LOF greater than 1 indicates a potentially abnormal observation.
Local density
Measure of the concentration of data points in a specific neighborhood around a point of interest. It assesses how many observations are located in a given region of the space.
Reachability distance
The maximum distance between a point and its k-nearest neighbors, used to normalize distances and reduce the impact of local variations. It allows for comparing densities between different regions.
Density factor
Ratio between the local density of a point and the average density of its neighbors, quantifying the relative isolation of an observation. High values indicate points in less dense areas.
Local outlierness
Quantitative measure of the degree of anomaly of a point based on its relative position to surrounding points. It captures subtle deviations in regions of variable density.
LOF score
Numerical value resulting from the LOF algorithm indicating the level of anomaly of an observation. A score close to 1 suggests a normal observation, while high scores signal outliers.
Local depth
Concept measuring how deeply a point is integrated into a local data distribution. It influences anomaly detection in complex data structures.
Density-based isolation
Principle that anomalies are naturally more isolated in low-density regions. LOF exploits this property to identify atypical observations.
Density cluster
Grouping of data points forming a high-density region in the feature space. LOF uses these clusters as a reference to identify isolated points.
Decision boundary
Discriminant threshold separating normal observations from anomalies based on LOF scores. It can be adaptive depending on the local data distribution.
Parameter sensitivity
Property of LOF regarding its responsiveness to variations in parameter k and the choice of distance metric. It affects the ability to detect different types of anomalies.
Distance normalization
Process of adjusting distance measures to handle variable scales between different data dimensions. It improves the robustness of anomaly detection.
Similarity metric
Mathematical function quantifying proximity between two points in the feature space. It directly influences neighborhood formation and LOF scores.
Robustness to variations
Ability of LOF to maintain its performance in the face of changes in data distribution or noise. It depends on the adaptability of the local density measure.
Reachability density
Inverse of the average reachability distance of a point with respect to its neighbors, serving as a stable estimator of local density. It reduces the impact of statistical fluctuations.
Noise vs Anomaly
Distinction between random variations (noise) and significantly deviating observations (anomalies). LOF differentiates these phenomena through local density analysis.