Metody oceny wyniku grupowania w analizie skupień
Evaluation methods for the grouping result in cluster analysis

Summary/Abstract: The paper reviews evaluation methods for the grouping result. It discusses three methods how to choose the correct number of clusters for agglomeration methods proposed by Mojena and Wishart, which are implemented in ClustanGraphics 8. The criteria which are mentioned are based on the relative sizes of the different fusion levels of the objects in the dendrogram – best cut significance test (upper tail, moving average) and checking randomness of objects clustering in the dendrogram – validation tree. The text is followed by an empirical example.

• Details
• Contents