Knowledge Graph-based Workflow Validation Using PROV Graphs
Abstract
Workflow execution validation is typically performed directly on event logs using procedural checks that become increasingly difficult to maintain as validation rules involve richer relationships among activities, resources, cases, and execution histories. This paper investigates knowledge graphs as an alternative representation for workflow execution validation. We present a lightweight approach that transforms XES event logs into PROV-aligned knowledge graphs and validates workflow executions through SPARQL queries on a triple store. The proposed representation makes workflow entities and their relationships explicit, enabling behavioral, role-aware, attributeaware, and provenance-aware validation within a unified graph-based model. We implemented a prototype and evaluated it using the BPI Challenge 2020 Domestic Declarations event log. The evaluation examines graph-construction and validation performance, rule effectiveness, and the trade-offs between graph-based and traditional log-based validation. The results show that the knowledge graph introduces additional preparation cost but provides a unified and inspectable representation for expressing workflow constraints, particularly those involving roles, provenance relations, and multi-step dependencies.