KI-Glossar
Das vollständige Wörterbuch der Künstlichen Intelligenz
Ward's Linkage
Aggregation method minimizing intra-cluster variance at each merge, producing compact and spherical clusters in the dendrogram.
Cut Height
Distance threshold on the vertical axis of the dendrogram determining the optimal number of clusters according to the chosen stopping criterion.
Agglomerative Coefficient
Index measuring the strength of the hierarchical clustering structure, ranging from 0 to 1, indicating the cohesion of merges.
Single Linkage
Method based on the minimum distance between pairs of elements from two different clusters, tendency to create elongated clusters.
Complete Linkage
Approach using the maximum distance between elements of clusters, favoring the formation of compact and well-separated groups.
Average Linkage
Strategy based on the average of distances between all pairs of elements from two clusters, offering a compromise between single and complete linkage.
Linkage Matrix
Data structure storing the sequence of merges and associated distances, serving as the basis for dendrogram construction.
Dendrogram Reordering
Process optimizing the order of leaves in the dendrogram to minimize crossings and improve visual interpretability.
Cluster Tree
Alternative hierarchical structure to the dendrogram representing parentage relationships between clusters at different levels of granularity.
Heatmap Dendrogram
Visualization combining a dendrogram with a heatmap to simultaneously explore hierarchical structures and data patterns.
Dynamic Tree Cutting
Adaptive algorithm determining clusters by identifying significant branches of the dendrogram based on their shape and depth.
Merge Sequence
Chronological order of cluster aggregations in the hierarchical process, stored in the linkage matrix for dendrogram reconstruction.
Cluster Distance
Dissimilarity measure between two groups of observations according to the chosen linkage method, determining successive mergers.
Dendrogram Elbow Criterion
Visual method identifying the inflection point in merge distances to determine the optimal number of clusters.
Dendrogram Bootstrap
Resampling technique evaluating the robustness of dendrogram branches by calculating p-values for each merger.
3D Dendrogram Visualization
Three-dimensional spatial representation of hierarchical structures enabling interactive exploration of complex relationships between clusters.