Determination of the Number of Clusters and Identification of Outliers in Statistical Software Packages

Abstrakt

The paper deals with possibilities how to determine the optimal number of groups of objects and find outlying objects when objects are clustered by different methods implemented in commercial statistical software packages. In the example, the aim is finding groups of similar binary variables. The methods as cluster analyses (hierarchical, k-medoids, fuzzy, two-step), multidimensional scaling, factor analysis and Boolean factor analysis are used.

Kategorie a štítky

  • Konferenční příspěvek (mezinárodní konf.)