KI-Glossar
Das vollständige Wörterbuch der Künstlichen Intelligenz
Distance to Centroid
Metric used to evaluate how far a point is from the center of its assigned cluster, where a high distance indicates a probability of anomaly.
K-Means Anomaly Score
Score calculated as the minimum distance of a point to all K-Means cluster centroids, where points with the highest scores are considered anomalies.
Cluster-Based Local Outlier Factor (CBLOF)
Method that evaluates the deviation of a point from its assigned cluster and the size of that cluster, considering as anomalies points far from large clusters or belonging to small clusters.
Silhouette Outlier Detection
Technique using the silhouette coefficient to identify anomalies, where points with very negative coefficients are likely misassigned and therefore abnormal.
Gaussian Mixture Model (GMM) Anomaly Detection
Probabilistic approach modeling data as a mixture of Gaussian distributions, where points with low probability under the model are identified as anomalies.
BIRCH (Balanced Iterative Reducing and Clustering using Hierarchies)
Efficient hierarchical clustering algorithm for large datasets, capable of identifying anomalies as points not fitting into the micro-clusters of the CF Tree.
Spectral Clustering for Anomaly Detection
Method using eigenvalues of the similarity matrix to perform clustering in reduced dimension space, where anomalies appear as isolated points in the spectrum.
Affinity Propagation Outlier Detection
Clustering technique based on message passing between points, where anomalies are identified as points never becoming exemplars or having low affinity with formed clusters.
Subspace Clustering for Anomalies
Approach performing clustering in different subspaces of dimensions to capture anomalies that are only detectable in specific projections of the data.
Agglomerative Hierarchical Clustering for Anomalies
Method building a hierarchy of clusters where anomalies are identified as points merged late in the dendrogram or forming isolated singletons.
Fuzzy Clustering (Fuzzy C-Means) for Anomalies
Variant of K-Means where each point has a degree of membership to each cluster, anomalies being characterized by low and uniformly distributed membership degrees.
Incremental Clustering for Anomaly Detection in Streams
Adaptation of clustering algorithms for continuous data streams, where anomalies are points that do not integrate into dynamically updated cluster models.
Robust Clustering to Anomalies
Family of clustering algorithms designed not to be influenced by outlier points, enabling better separation between normal clusters and anomalies.
Gravity-Based Clustering
Clustering method inspired by physics where anomalies are points that are not sufficiently attracted by the