Author(s): Vincent Marcov
With regards to gene expression data, several internal and external validity measures have been recommended to approximate the number of their clusters. Nonetheless, in most cases, the experts fail to take into account the analysis related to how a clustering algorithm can produce stability of the groupings . The article recommends a new selection process and standard of cluster validation to establish the appropriate number of clusters. The method evaluates the predictive stability or power of partitioning.