ABSTRACT
Machine Learning can only work with a lot of data, i.e. a lot of volumes, so that the algorithms can be trained on the part of the dataset in order to establish what is known as a ground truth, which will be validated by experts in the field. Making entities as vast as cities, economic sectors and populations comparable implies reducing their internal diversity a priori. Today, the availability of online traces of behaviors of all kinds makes it possible to detect patterns, for instance, from the activity of accounts on social media, based on likes, links, republication, etc., and no longer on public opinion. Comparisons are potentially innumerable when one goes down to this finely grained level of traceability. A specific suburb or place may have a career that can be tracked in the conversations, the reports, and the signals that all stakeholders select in a period of time.
