Читать книгу Data Analytics in Bioinformatics - Группа авторов - Страница 43

Hierarchical clustering technique is again divided into two categories

Agglomerative clustering—This follows a bottom up approach as depicted in Figure 2.4(a). Initially, every other object in the dataset is assumed as a single cluster. Distance measures are computed upon each object and based upon the similarities they are merged to form clusters.

Figure 2.4 (a) Agglomerative clustering, (b) divisive clustering.

Divisive Clustering—This follows top down approach as depicted in Figure 2.4(b). Initially, it considers all the data points as single cluster or model. The model separation is performed recursively until required criteria is met based upon the model.

Various algorithms of hierarchical clustering are described below.