Data Science Stack Exchange is a question and answer site for Data science professionals, Machine Learning specialists, and those interested in learning more about the field. It only takes a minute to sign up.
Sign up to join this community
It is known that the results of hierarchical agglomerative clustering using single-link method in order to determine the inter-cluster distance suffer form the chain-effect (natural clusters tend to extend trough a line of few points, like in the image below).
Does centroid-linkage have the same disadvantage?
The potential worst cases of centroid linkage are probably too crazy to explain as easy as the single-link effect... To see how it reacts to the classic Single-Link problem, why don't you just try yourself?
Roughly, because of the way centroid linkage works, it may end up using a virtual cluster center that is outside of the actual cluster. You then may see some very weird links to happen. That is also why it can have non-monotone linkage levels (so later merges may be cheaper than earlier merges).
