An ER-Model-Based Framework for Case Notion Selection in Object-Centric Processes

2026-07-28

AI Summary

This work addresses the limitations of traditional case notions in object-centric process mining, which either oversimplify coordination semantics through flattening or become overly complex due to resource objects. To overcome this, the study introduces entity-relationship (ER) modeling into case conceptualization for the first time. By identifying primary entities as process anchors and distinguishing secondary coordinating entities from resource entities, cases are automatically generated as connected components over the induced relationship graph. This approach yields a semantically sound and structurally concise partitioning that supports transitive closure propagation. Evaluated on an OCEL event log comprising 1,000 events, the method successfully produced 40 coherent and semantically clear cases, effectively enabling downstream process discovery and conformance analysis.

Abstract

Object-centric process mining operates on event logs where each event references multiple objects of different types. A fundamental challenge is defining a case notion - the grouping of events into coherent process execution instances -without which process discovery and conformance checking cannot proceed. Existing approaches either flatten the log to a single object type (losing inter-object coordination) or use the connected component of the object graph (creating overly complex cases due to resource-like objects). We propose an Entity-Relationship-schema-guided framework that identifies the primary entity (PE) type anchoring each process execution, classifies other entity types as secondary coordination entities or resources, and defines cases as connected components of the primary and secondary entity object relationship graph. The resulting case notion is shown to produce strict partitions with automatic transitive closure. The framework is illustrated with an order management process and validated on a 1,000-event OCEL log producing 40 structurally coherent cases.

Problem

• object-centric process mining
• case notion
• event logs
• process execution
• Entity-Relationship model

Innovation

• object-centric process mining
• case notion
• Entity-Relationship model
• primary entity
• connected component