engleski

Independent component analysis as a dimensionality reduction method on public sector buildings energy data

Excessive energy consumption affects large financial costs as well as the environment pollution, contributing to climate change and global warming. Croatia is among the highest ten energy intensity countries in European Union. Furthermore, since that 40% of all energy consumption in the European Union belongs to the building sector, this sector deserves attention. In that sense, the modelling of energy consumption is quite important topic. In this research, a real data set obtained from Croatian Energy Management Information System was used. It consisted of the constructional, geospatial, meteorological, occupational and energy data of public sector buildings in Croatia. It can be considered as a large data set, having a lot of observations and consisting of large number of variables. In this paper, the difficulties caused by the large number of variables in the model will be mentioned and the independent components analysis (ICA) as one method for dimensionality reduction will be discussed. ICA is a method for finding underlying factors or components from multivariate statistical data that looks for components that are both statistically independent and nongaussian. In this paper, the aforementioned method for reducing the dimensionality will be applied, as a step that will precede the modelling of the energy costs of public sector buildings depending on their characteristics.

dimensionality reduction ; feature extraction ; independent component analysis ; large data sets

nije evidentirano