Two commonly used normalization schemes are linear scaling to unit range (min–max normalization) and linear scaling to unit variance (z-score normalization). Several studies revealed that the former scheme is preferable over the latter since the latter is likely to eliminate valuable between-cluster variation