Automatic identification of the number of clusters in hierarchical clustering

Hierarchical clustering is one of the most suitable tools to discover underlying true structure a dataset in case unsupervised learning where ground truth unknown and classical machine classifiers are not suitable. In many real applications, it provides perspective on inner data preferred partitional methods. However, determining resulting number clusters hierarchical requires human expertise deduce this from dendrogram represents major challenge making fully automatic system such as ones required for decision support Industry 4.0. This research proposes general criterion perform cut automatically, by comparing six original criteria based Calinski-Harabasz index. The performance each 95 real-life dendrograms different topologies evaluated against classes proposed experts winner determined. framed bigger project build an Intelligent Decision Support assess 3D printers sensor real-time, although can be used other applications clustering.The methodology applied huge reduction CPU time also shown before after modification entire method. It reduces dependability human-expert provide inspecting dendrogram. Further, process allows applying mode industrial continuous monitoring production, helps building System detect operational modes, anomalies, behavioral patterns.