Hierarchical Groups and Their Limits
Some routes appear nested within broader regional groups.
Read the full lesson and open its transcript below. A narrated video will be added after production.
What you will learn
Some routes appear nested within broader regional groups.
A practical question
Some delivery routes naturally sit inside larger regions. Hierarchical clustering builds a tree of merges based on a chosen distance and linkage rule. Cut the tree at different heights and you get different groupings.
The useful idea
Single linkage can join two distant collections through a chain of close neighboring trips. Complete linkage favors tighter merged groups. Neither rule tells us by itself where the true business categories begin and end.
Work through the example
Compare the same routes under both linkage choices and inspect where a rural route joins the city group. Ask whether that change helps any planning task. A pretty dendrogram is a summary of a method and its inputs, not proof of a natural taxonomy. Hierarchical clustering shows a sequence of merges rather than only one final partition. With one linkage rule, a chain of moderately similar trips may connect distant cases; with another, groups may stay compact. Compare the same delivery data under both choices and inspect where meaningful groups join. A tree diagram can guide a discussion, but the height of a branch is not a ready-made business category. Validate proposed groups against conditions the dispatcher recognizes. Now check the idea on a concrete example: Compare where two delivery groups join under single versus complete linkage. A possible trap is worth naming: A dendrogram is not proof of a unique natural hierarchy. The answer follows from that evidence: Hierarchical clustering builds a merge tree based on distance and linkage choices. A long chain of routes can connect two regions under one linkage rule. Look at the actual trips where the groups first join, then compare with a compact-group rule. If the business interpretation changes completely, the hierarchy needs cautious explanation.
Limits and next step
Next, we'll compress correlated features to a few directions for exploration.
Worked example
Compare where two delivery groups join under single versus complete linkage.
Quick check: test your understanding
Try it yourself: Consider this case: Compare where two delivery groups join under single versus complete linkage. Explain how you would check the result. In your answer, address this limitation: A dendrogram is not proof of a unique natural hierarchy.
Check the reasoning
Hierarchical clustering builds a merge tree based on distance and linkage choices. A dendrogram is not proof of a unique natural hierarchy.
What is the main idea behind hierarchical groups and their limits?
Hierarchical clustering builds a merge tree based on distance and linkage choices.
How is it used in the example?
Compare where two delivery groups join under single versus complete linkage.
What limitation should be checked?
A dendrogram is not proof of a unique natural hierarchy.
Further reading
Machine Learning — Tom M. Mitchell.
Original ToolsInVision teaching examples. Editorial approach · Suggest a correction
Accessible referenceLesson 59 transcript290 words · Narration ready for the upcoming videoOpen transcript
Some delivery routes naturally sit inside larger regions. Hierarchical clustering builds a tree of merges based on a chosen distance and linkage rule. Cut the tree at different heights and you get different groupings.
Single linkage can join two distant collections through a chain of close neighboring trips. Complete linkage favors tighter merged groups. Neither rule tells us by itself where the true business categories begin and end.
Compare the same routes under both linkage choices and inspect where a rural route joins the city group. Ask whether that change helps any planning task. A pretty dendrogram is a summary of a method and its inputs, not proof of a natural taxonomy. Hierarchical clustering shows a sequence of merges rather than only one final partition. With one linkage rule, a chain of moderately similar trips may connect distant cases; with another, groups may stay compact. Compare the same delivery data under both choices and inspect where meaningful groups join. A tree diagram can guide a discussion, but the height of a branch is not a ready-made business category. Validate proposed groups against conditions the dispatcher recognizes. Now check the idea on a concrete example: Compare where two delivery groups join under single versus complete linkage. A possible trap is worth naming: A dendrogram is not proof of a unique natural hierarchy. The answer follows from that evidence: Hierarchical clustering builds a merge tree based on distance and linkage choices. A long chain of routes can connect two regions under one linkage rule. Look at the actual trips where the groups first join, then compare with a compact-group rule. If the business interpretation changes completely, the hierarchy needs cautious explanation.
Next, we'll compress correlated features to a few directions for exploration.