Use of Methods of Multidimensional Grouping for Statistical Analysis of Energy Consumption at Regional Level

Apply

Use of Methods of Multidimensional Grouping for Statistical Analysis of Energy Consumption at Regional Level

Journal Title: Статистика України - Year 2015, Vol 70, Issue 3

Abstract

Effective energy consumption at macro- or mesolevel is an important factor of economic development Consumption of energy resources (coal, gas and electricity in case of this study) in Ukrainian regions is uneven, being dependent on factors like share of industrial and agricultural output in regional value added, average income per capita in a region, share of out-dated industrial technologies in a region. In view of this, the objective of the study is to make clustering of Ukrainian regions by coal, gas and electricity consumption. The analysis rests on testing the following hypothesis: due to geographic and economic differences of the regions, the consumption of the above resources is distributed by them unevenly, which leads to the occurrence of homogenous groups of regions by measures of consumption of these resources. The performed visualization of input data for the three variables (coal, gas and electricity consumptions) and testing of the results for the existence of a natural segmentation at the initial phase allows for the following findings: (i) the occurrence of two significant surges for consumptions of the three categories of resources, corresponding to Donetsk and Dnipropet.rovsk regions; (ii) the occurrence of at least two natural segments, caused by a wide gap in consumption of the two categories of resources, gas and electricity This leads to the assumption of the existence of two or three segments (clusters, homogenous groups) of Ukrainian regions by the three studied variables. Two different although well known algorithms for cluster analysis are taken, to group the regions by electricity consumption: the algorithm based on agglomerative hierarchical clustering by the method of “most distant neighbor”, and the algorithm based on divisional segmentation by the method of -averages. Both algorithms show nearly the same results, breaking the regions into the three clusters: regions with low consumption of energy resources (16 objects), regions with high consumption of energy resources (2 objects), regions with medium consumption of energy resources (9 objects). The following conclusions are made from the study: Ukrainian regions are not supposed to be homogenous from energy consumption perspective. When the task of constructing models for forecasting and planning of energy-related expenditures by region is set, it requires consideration for the results of clustering, with models constructed for each segment, in order to increase the accuracy and the reliability of forecasts. It should be noted that the main discriminating factors for regions are gas and electricity consumptions.

Authors and Affiliations

L. L. Poltavets

EP ID EP564317
DOI -
Views 154
Downloads 0