- Ward: The Ward linkage forms successive clusters by finding the clusters that minimize the increase in within-cluster variance at each step. It is the default option in sklearn’s hierarchical clustering API.
- Single: The next cluster is formed by merging the observation that has the shortest minimum distance to an existing cluster.
- Complete: The next cluster is formed by merging the observation that has the shortest maximum distance to an existing cluster.
- Average: The next cluster is formed by merging the observation with the shortest average distance to an existing cluster.
Note: In divisive clustering, clusters are formed by dividing existing clusters that maximize the choice of linkage, essentially working inversely to what was described for each option.