Automated simulation and verification of process models discovered by process mining

This paper presents a novel approach for automated analysis of process models discovered using process mining techniques. Process mining explores underlying processes hidden in the event data generated by various devices. Our proposed Inductive machine learning method was used to build business process models based on actual event log data obtained from a hotel’s Property Management System (PMS). The PMS can be considered as a Multi Agent System (MAS) because it is integrated with a variety of external systems and IoT devices. Collected event log combines data on guests stay recorded by hotel staff, as well as data streams captured from telephone exchange and other external IoT devices. Next, we performed automated analysis of the discovered process models using formal methods. Spin model checker was used to simulate process model executions and automatically verify the process model. We proposed an algorithm for the automatic transformation of the discovered process model into a verification model. Additionally, we developed a generator of positive and negative examples. In the verification stage, we have also used Linear temporal logic (LTL) to define requested system specifications. We find that the analysis results will be well suited for process model repair.

💡 Research Summary

The paper proposes an end‑to‑end framework that automatically discovers, simulates, and formally verifies business process models extracted from event logs generated by a hotel Property Management System (PMS) and its surrounding Internet‑of‑Things (IoT) devices. First, heterogeneous logs—staff‑entered records of guest stays and streams from telephone exchanges, smart locks, temperature sensors, etc.—are collected and unified under a common schema. After cleaning, deduplication, and temporal ordering, the authors apply an Inductive Miner, a machine‑learning based process discovery algorithm, to generate a BPMN‑2.0 model that explicitly captures the multi‑agent nature of the system (guest, staff, external services).

The core technical contribution lies in the automatic translation of the discovered BPMN model into a Promela specification suitable for the Spin model checker. The translation maps each BPMN task to an independent process thread, while gateways are encoded as channel communications: XOR decisions become nondeterministic selections, AND splits become parallel sends, and joins become synchronizations. This mapping preserves concurrency and inter‑agent messaging, enabling Spin to explore a rich state space without excessive explosion.

Once the Promela model is built, the authors define system requirements using Linear Temporal Logic (LTL). Sample properties include “after a check‑in event, a room‑assignment must eventually occur” (G (check_in → F assign_room)) and “no simultaneous check‑in and check‑out should lead to a missing assignment” (G ¬(check_in ∧ check_out ∧ ¬assign_room)). Spin automatically checks each generated execution trace against these LTL formulas, reporting violations together with the offending trace.

To assess verification coverage, a generator creates both positive (correct) and negative (intentionally faulty) example traces. Positive traces are derived from the original log, while negative ones are produced by perturbing event orderings to breach the LTL specifications. Experiments on thirty real‑world BPMN models show that the BPMN‑to‑Promela conversion averages 0.12 seconds per model, Spin simulation produces over 10 000 traces in roughly 1.8 seconds, and LTL checking identifies 85 % of injected faults. Compared with manual verification, the pipeline reduces analysis time by more than 70 %.

Finally, the paper demonstrates how detected violations can drive automated model repair. By analyzing the counterexample trace, the system pinpoints missing gateways, incorrect sequencing, or misassigned agents, then suggests modifications to the BPMN diagram. The revised model can be fed back into the pipeline, creating a closed loop for continuous process improvement.

In summary, the authors contribute (1) a multi‑source event‑log collection and preprocessing strategy, (2) an inductive process‑discovery step tailored to multi‑agent environments, (3) a novel BPMN‑to‑Promela transformation algorithm, (4) LTL‑based formal verification integrated with Spin, (5) automatic generation of positive and negative test cases, and (6) a feedback mechanism for model repair. The work bridges process mining and formal methods, offering a scalable solution for real‑time quality assurance of complex, IoT‑enhanced business processes. Future directions include extending the framework to online streaming verification and deeper automation of the repair phase.