Figure 1

How active are different regions and how focused are their recipe/ingredient preferences? This figure shows the normalized entropy of the recipe and ingredient frequency vectors per region versus the number of visits per region on a log scale. Low entropy means that there is a high regional focus on just a few recipes/ingredients while a large entropy indicates a more even distribution. We see that ingredient preferences are more focused than recipe preferences according to their lower entropy values; however, they are strongly correlated (Kendall $\tau =0.95$ for the list of regions ranked by their recipe entropy and ingredient entropy). Furthermore, we can observe that the most active regions are in Austria (due to their higher number of clicks) and that the more active a region is, the less random their recipe and ingredient preferences are (according to their lower entropy values).