A clustering technique that begins with the assignment of the number of clusters to be found. Points that will represent the centroids of these clusters are then evenly dispersed through the data and moved as if by gravity until they settle into positions in the data clouds and cease to move. This technique is much faster than the hierarchical technique but not as accurate, and is often used in electrofacies analysis when large data sets must be analyzed.

Related Terms:

cluster analysis
hierarchical cluster analysis