Classification of Energy Consumption in the Balkans using Ensemble Learning Methods

Abstract

This paper explores the use of ensemble learning methods for classifying the energy consumption from (i) gross domestic product, (ii) CO2 emissions, and (iii) total number of population. The proposed analysis extends the previous research where prediction of the energy consumption values was performed using regression strategies. The experiment involves energy use data from five different Balkan countries, which is collected in the period 1995-2014. The energy consumption classes are obtained from the energy use values by K-Medians and analysed. Then, multiclass ensembles of support vector machine and linear discriminant analysis are used for classification of the energy consumption given the gross domestic product, the CO2 emissions, and the total number of population. The classification task is compared with a traditional multiclass support vector machine approach. The obtained results are very promising.