Ocena ilościowego spożycia żywności przez rodziny polskie z wykorzystaniem ich klasyfikacji przy użyciu sieci neuronowych

Data from surveys carried out by GUS permit to state that there is large individual variability in consumption of food by the Polish population. However, traditional partitioning of households, for example, according to income or dwelling place does not fully explain this variability. It was assumed that neural networks as a new research tools can shade more light is respect to consumption of food in concentrated households. The use of neural networks of the Kohonen type enabled the separation of 4 basic concentrations of households differing in the level of food consumption as well as in the contribution of various products. This partitioning explained at a high level the variability in food consumption (the average correlations coefficient of groups of products examined reached the level of 0.6). Moreover this partitioning gave the opportunity to further analyze relatively homogenous households that clarity differ with each other.