What are the two ways in which Hierarchical clustering can proceed?

  • Agglomerative: In the agglomerative approach, each observation starts as its own cluster, and successive clusters are formed by combining existing clusters until all observations are part of one huge cluster (though in a practical use case, the researcher would identify a location in the hierarchy to define the final clusters).
  • Divisive: The divisive approach works opposite that of the agglomerative, as it starts with all observations in one large cluster and splits them into smaller clusters until theoretically each observation belongs to its own cluster. 

In each approach, hierarchical clustering merges (or divides) clusters using a specified linkage metric that defines how to determine the next cluster.