Development of a stopping rule of clustering performance by using the connected acyclic graph
Volodymyr Mosorov, Taras Panskyi, Sebastian Biedroń · Eastern-European Journal of Enterprise Technologies · 2015
In this article the technique of the analysis of a stopping rule for the data preclustering algorithm without the prior information about the number of clusters with the use of a connected acyclic graph is introduced. The connected acyclic graph (tree) makes it possible to represent the interconnection between the objects in input data. The stopping rule allows a halt at the some step assuming that further clusterization will not cause finding new clusters. The core of the analysis was the application of the preclustering algorithm and the stopping rule to the series of input data which were represented by sample cases of input data. Sample cases were input data with normal distribution law which belonged either to a single group or to many groups. The analysis has shown the advantages of the stopping rule for the data preclustering algorithm.