The Use of Combined Models in the Construction of Foodstuffs Consumption Forecasting in the Czech Republic

Many authors all over the world attempt to perform time series analyses (at differing levels of expertise) in their published works. Knowledge of quantitative information is necessary for decision making in any domain. Therefore, it is more desirable to enter this field of problems and examine and develop everything that has been offered by these modern methodologies. In time series forecasting, the extrapolation methods are applied most frequently in practice. Currently, the combined models have been increasingly employed in experiments – these represent an aggregation of prognoses obtained from various separate models. The study presented is aimed at such new approaches, i.e. the construction of combined prediction models that are more realistic, more flexible and more concise in the time series modelling. This paper focuses on a subsequent assessment of combined prognoses constructed and a comparison of these with selected separate models having participated in the aggregate prognoses making. In order to obtain an efficient product, the Time Series Forecasting System (TSFS) component has been employed, being a component of the SAS programme system. For quality assessment of the models constructed, the assessment criteria selected in advance have been applied. The results of this empirical study have shown that in the domain of estimation of future foodstuffs consumption development, the techniques illustrated in this paper by examples of long-term time series from foodstuffs consumption area in the Czech Republic (CR), can be employed with success. This way represents a suitable supplement to complex econometric models.