Characterizing the imbalance of Customs dataset-derived nutrition estimates through entropy distributions

Abstract

The nutrient distribution of food consumed on a regular basis has known impacts on health. In this work, we look at the imbalance in nutrient estimates for generated food plates from imports data by condensing the nutrient distribution information into an entropy metric describing the evenness of a plate's nutrient distribution. To examine which nutrients skew the entropy distribution of plates, we computed the Jensen-Shannon distance between the entropy distribution for N nutrients and the entropy distribution for N − 1 nutrients. A higher JS distance means that the removal of a nutrient skews the entropy distribution, which is a consequence of the change in the shape of the nutrient distribution. We found that dietary fiber poses an outsized role in changing nutrient distributions.