Agglomerative clustering algorithm that preserves ordering of the data points?

Question

I need a hierarchical clustering implementation (preferably in matlab) that preserves the order of my data points. For example, I don't want it to output something like this

(((1, 5), (2, 3)), 4)

I pretty sure that there is a better term for this. I just need it to combine the clusters are near each other in the same order as the input.

Answer 1

What you are trying to do is no longer cluster-analysis. It's similar, but not clustering.

You are trying to segment your data, i.e. you actually have a given sequence and look for good points to split it.

In the usual way of thinking in cluster analysis, you have lots of individual items, and you want to collect these so that nearby objects are aggregated; in a way that they may or may not be contiguous.

Really, instead of trying to hack a clustering algorithm to achieve your goals, it makes just tons more sense to start from your starting point, and just maybe "reuse" some of the ideas that were previously used in clustering.

The reason is not just technical, but also a matter of intuition.