Modeling and Discovering Dynamic Identity Relations in Object-Centric Process Mining

Modern information systems track the behavior of business objects in digital execution records, called objectcentric event logs. An important step in the analysis of these logs is the discovery of a process model that describes the interacting control flows of all objects. In real-life processes, these interactions are often based on the object’s identities. In an order management process, for example, the same set of items that is ordered by a customer is subsequently also delivered. However, existing object-centric modeling formalisms and discovery techniques either ignore these identity relations, treat them as static, or capture them without guaranteeing important behavioral soundness properties. In this paper, we formalize three types of dynamic identity relations that can frequently be found in real-life processes. Then, we introduce a new object-centric modeling formalism that supports these identity relations while guaranteeing important soundness properties by construction. Additionally, we provide a discovery algorithm to construct corresponding process models from object-centric event logs. We prove that the inclusion of dynamic identity relations in our models preserves and improves important model quality criteria. Lastly, we evaluate our approach by applying it to a range of public logs. We observe the discovery of dynamic identity relations in feasible runtime in all investigated logs.