Normally data analytics is done on databases through the use of data warehouses, I am wondering what might be different and insightful when one does the same but on an ontology.
I assume that you speak about OWL ontologies. With respect to a relational database schema, it offers useful features:
- the inheritance hierarchy is useful to gather data that would otherwise be in unrelated tables;
- it is useful to aggregate (align) data from different databases; they often have different coverage (scope) and this is reflected e.g. in the different attributes and tables.
1) the open world assumption of OWL permits to deal with missing information;
2) the use of a TBOX explicits the links between their data, and is the basis for automated reasoning (e.g. by Pellet or Racer).