Determination and validation of group numbers in a cluster analysis: A case study applied to forestry science

摘要

The objective of this study was to systematically study and interpret the number of species groups through the adjusted Rand index (ARI), from the combination of the hierarchical technique and connection method characteristics. We attempted to facilitate the work of researchers on the number of groups and validation of the groups. The original data based on the Mahalanobis distance, the hierarchical single linkage, complete linkage, average linkage, Ward’s and ARI methods were applied. For the validation of methods, cophenetic coefficients and the ARI were used. It was observed that the use of the ARI had the potential to significantly improve the decisions regarding the number operating groups, becoming, therefore, a recommended technique to search for the optimal number of groups.