Agglomerative hierarchical clustering : Flat clustering returns an unstructured set of clusters. On the other hand, this structure is more informative. We don't have to define the number of clusters in advance with this clustering procedure. Bottom-up algorithms start by treating each PIECE of data as a singleton cluster, then AGGLOMERATE pairs of clusters until all of them are merged into a single cluster that contains all of the data.

Divisive Clustering : This approach also eliminates the need to define the number of clusters ahead of time. It necessitates a method for breaking a cluster that contains all of the data and then recursively splitting clusters until all of the data has been split into singletons.

Following are the differences between the two :

• When compared to agglomerative clustering, divisive clustering is more complicated since we require a flat clustering algorithm as a "SUBROUTINE" to split each cluster until each data has its own singleton cluster.
• If we don't create a complete HIERARCHY all the way down to individual data leaves, divisive clustering is more efficient.
• A divisive algorithm is also more precise. Without first examining the global distribution of data, agglomerative clustering makes judgments based on local patterns or neighbour points. These early DECISIONS are irreversible. When generating top-level dividing decisions, divisive clustering takes into account the global distribution of data.