Characterisation of Clusters: Homogeneity, Separability

This lecture covers the concepts of centroid and medoid, homogeneity, and separability in clustering. It explains how to calculate homogeneity and separability, illustrating with examples of high and low homogeneity and separability. The lecture also introduces the Davies-Bouldin index and silhouette coefficient for evaluating clustering quality, as well as cluster stability and the Rand index. Additionally, it discusses expert knowledge in clustering, including ontology and enrichment analysis. Different clustering algorithms such as clustering by density, centroids, and hierarchical clustering are presented, emphasizing the importance of homogeneity and separability.