THE IMPACT OF THE AGGREGATION LEVEL ON THE FORECASTING ACCURACY OF PETROLEUM CONSUMPTION

Hierarchical forecasting represents an approach to forecasting which takes account of the possibility of integration of foreseen items into related groups in accordance with chosen aggregation variables. Therefore, it is possible to use it in creation of a system of forecasts on different aggregation levels in the way the created forecasts are in mutual time and material harmony. The thing is that in hierarchical forecasting the final forecasts are created from forecasts on a lower level of aggregation (bottom-up process), or from forecasts on a higher level of aggregation (top-down process). However, the knowledge of the influence of the aggregation level on the accuracy of the arising forecasts is a necessary condition of using the right approach to hierarchical forecasting. This paper describes the principles of hierarchical forecasting and the possibilities of their utilization within forecasting development of time series in consumption of basic petroleum products in the EU countries in 1993-2010. On the basis of the time series analysis, the authors then deal with comparison of the accuracy of forecasts arising on different levels of aggregation and evaluate the impacts of the aggregation level on the forecasting accuracy. Target literature search into scientific literature and a time series analysis of petroleum consumption in the EU are used as the research methods and sources.