SENSITIVITY OF THE EUROPEAN UNION REGIONAL LABOUR MARKETS TO ECONOMIC CRISIS. CLASSIFICATION USING WARD METHOD WITH COHESION CRITERION

The popular clustering procedures applied in administrative units clustering (regions, districts, municipalities) frequently result in distinguishing incoherent groups, which is difficult to accept by the analyses addressees. The intuitive approach also imposes the investigation of coherent divisions, since it greatly facilitates both the interpretation of results and their practical application. The study presents the modification of Ward method by applying the condition of cluster coherence at every agglomeration stage. Therefore the starting point of the clustering procedure is not only the distance matrix between objects, but also their mutual neighbourhood matrix, most often understood literally, in the geographical sense. While searching for two closest clusters only the neighbouring ones were considered. The modified clustering algorithm was implemented as the extension of STATISTICA software. The modified clustering algorithm is illustrated by the classification results of the European Union regions regarding sensitivity to economic crisis. The objective of analysis was to distinguish the coherent division of EU regions against the selected diagnostic variables illustrating changes on the EU labour markets in the period 2008-2012.