ABSTRACT
This chapter introduces the concepts of data scaling and its application. It depicts some classical clustering methods, k-means (and k-median), k-medoids, and agglomerative hierarchical clustering. Many statistical methods are deeply dependent on the ranges of distinct parameters of interest. The power transformation scaling methods help model skewed data sets, data sets with high variability, or circumstances where normality is desired. Clustering methods often require quantifying the distance between groups of items. Many authors have tried to formalize the definition of a cluster. However, it is a daunting task since their efforts led to numerous and diverse criteria. The methods based on partitioning organize objects into clusters, where each cluster is formed to optimize a partitioning criterion based on similarity among objects within a cluster and distinction between objects of different clusters.
