The assignment of a set of observations into subsets (called clusters) so that observations in the same cluster are similar in some sense (Wikipedia)