A Knowledge-Driven Approach to Interpretable Compliance Deviation Analysis in Business Processes

Event logs of enterprise information systems capture rich knowledge regarding business process execution, yet extracting interpretable business semantics for compliance monitoring remains challenging. This study proposes a four-layer deviation analysis framework based on process mining. At its core lies a formalized knowledge base, which incorporates designed deviation mapping rules and classification algorithms to systematically transform abstract technical semantics into three explicit categories of business deviations—missing, out-of-order, and redundant—thereby generating deviation semantics that can be intuitively understood by business personnel. The knowledge-driven framework has been implemented as a plug-in for the ProM platform. Experiments on both synthetic and real-life logs, along with expert evaluations, demonstrate that the proposed method exhibits strong generalization capability and effective deviation detection performance in complex scenarios, significantly improving diagnostic efficiency. Furthermore, by establishing a correlation mapping between deviations and fraud risks, the study closes the audit loop from deviation identification to risk localization. This offers a knowledge-driven practical solution for compliance control and risk prevention in enterprise process management based on information systems.