Abstract
In the field of Business Process Management, the Process Discovery task is one of the most important and researched topics. It aims to automatically learn process models starting from a given set of logged execution traces. The majority of the approaches employ procedural languages for describing the discovered models, but declarative languages have been proposed as well. In the latter category there is the Declare language, based on the notion of constraint, and equipped with a formal semantics on LTLf. Also, quite common in the field is to consider the log as a set of positive examples only, but some recent approaches pointed out that a binary classification task (with positive and negative examples) might provide better outcomes. In this paper, we discuss our preliminary work on the adaptation of some existing algorithms for Inductive Logic Programming, to the specific setting of Process Discovery: in particular, we adopt the Declare language with its formal semantics, and the perspective of a binary classification task (i.e., with positive and negative examples).