What is 2D dataset with two real groups of data where an agglomerative clustering method will be able to cluster the data correctly, and K-means will not?
I thought about a data set that would look like the figure below, but still I can't think of any real world features that would graph such a data set.