Cluster analysis is concerned with detecting relatively distinct data groups consisting of similar observations. The need in such data analysis has been seen in various areas of science. There are many clustering methods proposed in the literature that can be divided into two general groups: hierarchical methods with different merging or splitting rules called linkages (Ward, 1963) and partition-optimization algorithms such as K-means (Forgy, 1965 and MacQueen, 1967) and K-medoids (Kaufman and Rousseuw, 1990). In this paper, we discuss some variations of the most well-known partition-optimization algorithm—K-means.