Synonyms

Agglomerative hierarchical clustering; Agglomerative hierarchical data segmentation; Bottom-up hierarchical clustering

Definition

Cluster analysis consists to classify a set of objects (observations, individuals, cases) into subsets, called clusters, such that they have similar characteristics or properties. There are different ways to define the similarities among objects or variables through the use of different metrics. Some of them are as follows:

  • The single-linkage clustering, or nearest neighbor clustering, takes into account the shortest distance of the distances between the elements of each cluster. This is one of the simplest methods.

  • The complete linkage clustering, or farthest neighbor clustering, takes the longest distance between the elements of each cluster.

  • The average linkage clustering takes the mean of the distances between the elements of each cluster. The merged clusters are the ones with the minimum mean distance.

There are a variety of clustering algorithms; one of them is the agglomerative hierarchical clustering. This clustering method helps us to represent graphically the results through a dendogram. The dendogram has a tree structure that consists of the root and the leaves; the root is the cluster that has all the observations, and the leaves are individual observations. The agglomerative hierarchical clustering starts with the individual observations and successively fuses the clusters that are closer together (the most similar ones).

Cross-References

Modules, Identification Methods and Biological Function