Finding and comparing even small subgroups in large structured datasets

What is it about?

Imagine a room of 100 persons. If you start looking for specific attributes of these persons, e.g., must be younger than 25 and wear a green hat, there might be only a few left or even none. Without knowing these persons in advance, it is quite difficult to estimate the size of these subgroups. If, in addition, to the persons you also know what items each person bought in their last visit to the grocery store, then you might wonder if specific subgroups prefer certain products that they bought. Our approach allows you to see all of these correlations in a single condensed picture, allowing you to compare subgroups easily and spot subgroups that show a distinct behavior.

Featured Image