Hey Alessandro, I too have a similar use case where the cardinality of my variables is very high in case of categorical attributes.How did you solve this with the limited number of features?