Agglomerative Clustering is a type of hierarchical clustering that takes the bottom-up approach

Algorithm

agglomerative-clustering(points) {
	each point is its own cluster
	repeat until one cluster left:
		pick 2 "CLOSEST" clusters
		merge them into new cluster
	return that cluster
}

produces a family of clusterings represented by a dendrogram

Defining “Closest” for Clusters of Multiple Elements

  • closest pair (single-link clustering)
  • farthest pair (complete-link clustering)
  • average of all pairs
  • Ward’s Method (min variance, likek-means) - find pair of clusters that leads to minimum increase in total within cluster distance after merging