A
statistical technique for figuring out how to
clump things together. This seems like an easy
problem which you could just solve
visually, but sometimes the
clusters are not obvious, especially if the
elements you want to cluster are
scattered over a large area.
There are a number of different variations on this technique, but they all essentially work the same way. First, find the two elements which are closest together, and make a new, single element out of them. That's one cluster. Now, repeat. Repeat again until everything's assigned to a cluster, or you have the right number of clusters (decided a priori), or whatever. The variation comes in deciding the position of the new cluster you create. You can make it the position of the first point you chose, for instance, or the average position of all the original points in the cluster, or the average position of the new point and the current position, and so forth. These subtle differences can lead, in some cases, to dramtically different sets of clusters.