Skip to content
ML Atlas

Hierarchical Clustering

Builds a tree of clusters by repeatedly merging the two closest groups, so you can cut the tree at any level to get as many clusters as you like.

Also agglomerative clustering · dendrogram · linkage · Ward's method

Fig. / How it works
MERGE THE CLOSEST GROUPS
Role · kind
Models and algorithmsModel
Level
Intermediate

Start with every person as their own family. Join the two most similar people, then the next most similar pair or group, and so on, until everyone is one big family tree.

Where it fits

ClusteringClusteringHierarchical Clustering
+All relationships as a list

Is a kind of / part of

See it on the full map

How it works

Instead of producing one flat set of groups, hierarchical clustering produces a whole family tree of groups, called a dendrogram.

The common, bottom-up (agglomerative) version:

  1. Start with each point as its own cluster.
  2. Find the two closest clusters and merge them.
  3. Repeat until everything is in one cluster.

“Closest” between two groups needs a definition, called the linkage:

  • Single: distance between the two nearest members. Can form long chains.
  • Complete: distance between the two farthest members. Gives compact clusters.
  • Average: the mean distance between all pairs.
  • Ward: merge the pair that increases within-cluster variance the least, similar to what k-means optimizes.

To get, say, five clusters, cut the tree at the height where it has five branches. Long vertical lines in the dendrogram suggest natural places to cut.

It is ideal for exploring smaller datasets with nested structure. For millions of points, k-means or DBSCAN scale better.

Everyday use cases

  • Case 01

    Biology

    Building family trees of species or grouping genes with similar expression patterns into heatmaps.

  • Case 02

    Organizing documents

    Grouping news articles into topics and sub-topics without choosing a number of topics upfront.

Strengths, weaknesses and when to use it

Strengths

  • No need to fix the number of clusters in advance; the dendrogram shows structure at every scale.
  • Works with any distance measure.

Weaknesses

  • Slow and memory-hungry for large datasets.
  • Merges are greedy and never undone; results depend strongly on the linkage choice.

When to use itFor small to medium datasets where nested structure matters, or to explore how many clusters make sense.

Compare with

Check yourself

  1. Q01

    What is a dendrogram?

    Show answer

    A: A tree diagram showing the order in which clusters merged

Go deeper

Sources

  1. [1]scikit-learn User Guide: Hierarchical clustering