Pith. sign in

REVIEW 4 major objections 5 minor 50 references

Synchronizing Process Model and Event Abstraction for Grounded Process Intelligence (Extended Version)

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper proves that, under defined restrictions on event logs, abstracting a process model and abstracting its event log can be synchronized so that rediscovery from the abstracted log yields the same model as direct model abstraction.

desk verdict Genuine formal result, but the event abstraction transposes observed traces into canonical orders that may not occur in the log, undercutting the 'grounded' claim; still worth peer review. read the letter →

arxiv 2505.23536 v1 pith:BAFILMFL submitted 2025-05-29 cs.AI

Janik-Vasily Benzin , Gyunam Park , Stefanie Rinderle-Ma This is my paper
classification cs.AI
keywords processminingmodelabstractioneventbehavioralprofilessynchronizationdiscoveryInductiveMinerdirectly-followscompleteness
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish a synchronization theorem for process intelligence: when a discovered process model is abstracted by behavioral-profile abstraction, the underlying event log can be abstracted in a matching way so that rediscovering a model from the abstracted log yields a model isomorphic to the directly abstracted one. The motivation is practical: model abstraction alone produces a smaller model but leaves the event log at the original granularity, so downstream analyses lose their grounding in observed behavior and cannot be run on the simplified model. The paper proves the theorem for the Inductive Miner and for a non-order-preserving abstraction technique, and it designs a concrete event abstraction $eabpa$ that maintains trace frequencies while making the abstracted log directly-follows complete. The guarantee is conditional: the log must be `restricted` in a precise sense, and the paper includes counterexamples showing that synchronization can fail when those restrictions are relaxed.

What carries the argument

The load-bearing object is the synchronization square, the requirement that $pd(ea(L)) \cong ma(pd(L))$ for model abstraction $ma$, event abstraction $ea$, and discovery $pd$. The proof is carried by four components. The class $C_a$ of process trees (no duplicate activities, only single-activity self-loops, no silent steps outside self-loops) is the class that $mabpa$ always produces and that the Inductive Miner can rediscover isomorphically from any directly-follows complete log. The $ntl$ algorithm computes the number and lengths of traces in the minimal directly-follows complete log of any such tree, giving the reference log that $eabpa$ matches. The transposition equivalence relation, built from the Kendall-tau sequence distance, lets $eabpa$ split preliminary abstracted traces evenly across reference trace classes and transpose them into the correct order. Finally, the restricted-log definition (Definition 5) is what guarantees the abstracted minimal log is never larger than the concrete log, so an event abstraction that synchronizes can exist.

What would settle it

Take a log that violates Definition 5 but is otherwise simple, such as $L = \{\langle a,b\rangle, \langle e,c,d\rangle, \langle d,e,c\rangle\}$ from the paper's Activity Once Per Trace counterexample. Applying $mabpa$ with $agg(x)=\{a,b\}$ and $w_t \le 0.5$ yields an abstracted log with 7 traces while $L$ has 3, so no synchronized event abstraction exists, confirming the restriction is necessary. To test Theorem 1 itself, run $eabpa$ on any log that satisfies Definition 5 and check whether $pd_{IM}(L^a)$ is isomorphic to $mabpa(pd_{IM}(L))$; a single restricted log where they differ would refute the theorem.

Watch

Extended reading notes

Core claim

At the center is a commutativity statement. Let $L$ be an event log, $pd_{IM}$ the Inductive Miner, and $mabpa$ the behavioral-profile abstraction that clusters concrete activities into abstract activities from their weak-order relations. The paper proves Theorem 1: if $L$ is restricted and $mabpa$ is applicable to $M = pd_{IM}(L)$, then $pd_{IM}(L^a)$ discovered from the abstracted log $L^a = eabpa(L, pd_{IM}, mabpa)$ is isomorphic to $mabpa(pd_{IM}(L))$. The new event abstraction $eabpa$ works in two steps: first it replaces concrete events by abstract events trace by trace, deletes events that are in choice relation, and repeats an event when the abstract activity is in parallel with itself; then it transposes traces, using the Kendall-tau sequence distance, so that the result is directly-follows complete with respect to the abstract model while preserving trace multiplicities. The paper also proves the size condition that makes an event abstraction possible at all: for restricted logs, the abstracted minimal directly-follows complete log has strictly fewer traces and fewer events than the concrete log.

Load-bearing premise

The load-bearing premise is that the event log is `restricted` in the paper's precise sense: the Inductive Miner must only use the Strict Tau Loop and Empty Traces fall-throughs, only choice, parallel, and sequence cuts, and no sequence node may have an activity or self-loop child, so many real logs fall outside the guarantee and the paper's own counterexamples show synchronization can fail once these conditions are relaxed.

Editorial extensions

If this is right

  • After synchronization, process enhancement, simulation, and prediction can be run on the abstracted model together with the abstracted log, so those analyses remain grounded in observed trace frequencies and data values rather than in a log generated by replaying the abstracted model.
  • Because the second step of $eabpa$ evenly splits trace multiplicities across matched reference traces, the abstracted log preserves relative trace frequencies and value distributions to the extent the matching allows.
  • Any non-order-preserving model abstraction that hopes to be synchronizable must ensure the abstracted model's minimal directly-follows complete log is no larger than the concrete log; otherwise the paper's Lemma 6 shows no event abstraction can exist.
  • The theorem restricts the Inductive Miner to choice, parallel, and sequence cuts plus strict-tau-loop and empty-trace fall-throughs; extending the result to loop cuts or other fall-throughs would require widening the classes $C_c$ and $C_a$.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • One testable design rule suggested by the proof is that a model abstraction is synchronizable only if it never turns a sequence into a choice or parallel structure in a way that enlarges the minimal directly-follows complete log; measuring that log size before designing the event abstraction could serve as a screening test for other abstraction techniques.
  • The timestamp healing after transposition is only sketched, so a practical extension would define attribute-specific aggregation and timestamp repair operators with their own correctness conditions, since downstream prediction and simulation depend on timestamps and data values.
  • The restrictions exclude nested loops and duplicate activities, so an empirical question is how often real event logs satisfy Definition 5 or can be preprocessed to do so; the paper does not provide that empirical evidence.
  • Because the abstraction preserves trace multiplicities up to the chosen matching, one could compare the abstracted log's trace-frequency distribution against the concrete log's as an operational measure of how faithfully a given abstraction goal is met.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper formalizes a synchronization framework between model abstraction (MA) and event abstraction (EA) for process mining. It adapts behavioral profile abstraction (BPA, mabpa) and designs a two-step EA technique (eabpa) that first abstracts events trace-by-trace and then transposes traces to match a reference minimal df-complete log of the abstracted model. The main theorem (Theorem 1) states that, under a restrictive class of event logs (Definition 5), discovering a process model from the abstracted log yields a model isomorphic to the model obtained by direct abstraction of the originally discovered model. The extended version contains an appendix with proofs of the lemmas.

Significance. If the formal results are correct, the paper would provide the first synchronization guarantee for a non-order-preserving MA technique, enabling downstream process intelligence tasks on abstracted logs and models. The paper is explicit about its restrictions and provides counterexamples showing that synchronization can fail when they are relaxed. The formal framework is original and the appendix goes some way toward machine-checkable rigor, though several proofs are sketches. However, the practical claim of 'grounded process intelligence' is substantially overclaimed, because the constructed abstracted log is not a faithful projection of observed behavior at the trace level.

major comments (4)
  1. [Sect. 6.1 and Appendix A, Lemma 2] There is a direct inconsistency between the main-text Lemma 2 and the appendix version. In Sect. 6.1, Lemma 2 states that eabpa returns a df-complete log, and the proof shows La ⊆ Lr. In Appendix A, the same lemma is stated as returning a 'mdf-complete' log and the proof claims L1_a = La. The running example in Sect. 4.2 contradicts the appendix statement: there Lr has 46 traces while La has 4, so L1_a = La is false. This discrepancy must be resolved; if only df-completeness is intended, the appendix statement and proof need correction, and Theorem 1 still follows from df-completeness, but as written the appendix proof is invalid.
  2. [Sect. 6.1, Lemma 8] The proof of Lemma 8, which ensures the existence of a transposition matching between Ltmp and La, is not rigorous and is load-bearing for Lemma 2 and Theorem 1. The argument that |Ltmp/≈| = |La/≈| is made by contradiction relying on an informal claim about the number of choice nodes in the discovered model Mtmp, and it is asserted without proof that Ltmp satisfies the restrictions of Definition 5. The step 'From the last two statements, it follows that there exist two traces...' is not logically derived. This gap needs to be closed for the main theorem to be fully established.
  3. [Sect. 4.2 and Sect. 1] The claimed contribution of an EA technique that 'maintains observed distributions in the abstracted log' is not supported by the algorithm. In ea2_bpa (Algorithm 3), traces are transposed to match the canonical orders of the reference log Lm(Ma). In the running example, 30 of the 45 traces of the form ⟨AB,AC,FDD,FDD,SC,AP⟩ are transposed to orders that do not appear in the input log. The formal theorem only establishes df-completeness of Lr with respect to Ma, which any log with the same DFG would satisfy; it does not preserve trace-level order distributions or grounding in observed behavior. The paper's introductory and concluding claims of 'grounded process intelligence' should be weakened to model-level synchronization, or the transposition step needs to be justified as preserving the intended notion of grounding.
  4. [Appendix A, Lemma 1 proof] The proof of Lemma 1, a key well-definedness result, contains an explicit caveat in the sequence case: 'we do not consider optional nodes for simplicity', with an argument that ignoring them yields an upper bound on the abstracted log. This is a plausible but unproven assertion, and since Lemma 1 is used to guarantee |La| < |L|, the omission should be filled with a rigorous argument. Similarly, the proof of Lemma 7 relies on informal statements about 'code inspection' of the ordering-relation derivation and the alignment of its priority order with the operator ordering of Lemma 6; this should be made explicit.
minor comments (5)
  1. [Abstract] Typo: 'loosing' should be 'losing'.
  2. [Sect. 5.2, Algorithm 4] The notation in Lemma 5 ('LmpMq.log', 'lens') is not consistently defined; the algorithm returns both a count and a length sequence, but the lemma should state clearly how these correspond to the actual trace multiset.
  3. [Sect. 4.2, after Algorithm 3] The paragraph claiming that the 'perfect matching' uniquely exists is informal and would benefit from a precise statement or a reference to the formal proof in Lemma 8.
  4. [Appendix A, Lemma 2] The heading of the appendix lemma says 'mdf-complete' while the main text says 'df-complete'; this wording confusion should be fixed.
  5. [Conclusion] The paper would benefit from a discussion of how severe the restrictions in Definition 5 are for practical event logs, given that the paper's own counterexamples show immediate failure when any of the fall-through or cut conditions is relaxed.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Theorem 1 is a constructive synchronization proof, not a reduction to its own assumptions.

full rationale

The derivation chain is self-contained and built from independently established components. The paper defines synchronizability (Definition 1) as the existence of an EA technique making pd(eama(L)) isomorphic to ma(M), then explicitly constructs eabpa with the design goal that La be df-complete for Ma (Sect. 4.2). This design-to-target feature is a constructive guarantee, not circular reasoning: Algorithm 3 transposes traces of the preliminary abstracted log to match a reference minimal df-complete log Lm(Ma), and Lemma 2 then verifies df-completeness using Lemma 8 (matching of quotient sets) and Lemma 1 (size monotonicity). The rediscovery step is supported by Lemma 3, which extends the external Inductive Miner rediscoverability theorem of Leemans et al., and Lemma 4, which shows mabpa outputs trees in class Ca; neither lemma depends on the target result. Theorem 1 then combines Lemma 2 with Lemma 4 in the standard df-completeness-to-rediscovery argument. The restrictions in Definition 5 are stated assumptions, not hidden circular inputs, and the paper's own counterexamples show where synchronization can fail when they are relaxed. The transposition step altering event order is a faithfulness/grounding concern, not a circularity concern: it does not make the proof assume what it proves. No fitted parameter is renamed as a prediction, and no load-bearing self-citation or imported uniqueness claim is present. The novelty is the constructive EA design and its correctness proof under explicit restrictions, which remain independent of the theorem being derived.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

No fitted constants or newly postulated physical entities. The theorem is conditional on user-chosen aggregation parameters and on the strong restricted-log assumption. The main burden is the restricted-log condition and the reliance on unformalized, partially sketched lemmas in the appendix.

free parameters (2)
  • wt (order relation frequency threshold) = restricted to (0, wminmax]
    User-chosen threshold in mabpa; restriction (5) in Def. 3 bounds it to avoid primitive modules. The proof considers the maximal case wt = wminmax, but the parameter is not fitted to data.
  • agg (activity clustering function)
    User-defined mapping from abstract activities to sets of concrete activities; conditions (3)-(4) in Def. 3 restrict it. The paper says the actual value has no impact on the proof, but applicability of mabpa depends on it.
assumptions (5)
  • standard math IM is isomorphic rediscoverable from df-complete logs for process trees in classes Cc and Ca (extending Leemans et al., Theorem 14).
    Invoked in Lemma 3 and used in Theorem 1 to infer pd_IM(La) is isomorphic to Ma from df-completeness.
  • domain assumption Event log L is restricted (Def. 5): IM uses only 'Strict Tau Loop' and 'Empty Traces' fall-throughs, only choice/parallel/sequence cuts, and no node with an activity or self-loop child is a sequence node.
    This is the central scope condition of the theorem. Counterexamples in Sect. 5.3 show synchronization can fail when the condition is relaxed.
  • domain assumption mabpa is applicable to M (Def. 3): no duplicate activities, the MDT of the abstract profile has no primitive module, aggregation satisfies A_new/A_c conditions, and wt <= wminmax.
    These conditions guarantee that Ma is smaller, is in class Ca, and can be synthesized from the abstract behavioral profile.
  • standard math Language-preserving reduction rules and the normal form of process trees from [15] apply.
    Used in Lemma 1 to decompose abstracted trees and compare trace counts and sizes of mdf-complete logs.
  • standard math Kendall tau sequence distance is undefined for traces with different multisets, and transpositions can be applied to reorder traces.
    Defines transposition equivalence, which underpins the matching and even-split steps of ea2_bpa.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Synchronizing Process Model and Event Abstraction for Grounded Process Intelligence (Extended Version)." pith.science (2026). https://pith.science/paper/BAFILMFL

@misc{pith2026250523536,
  author       = {Pith},
  title        = {Pith review of: Synchronizing Process Model and Event Abstraction for Grounded Process Intelligence (Extended Version)},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BAFILMFL}},
  note         = {Machine review of arXiv:2505.23536}
}
read the original abstract

Model abstraction (MA) and event abstraction (EA) are means to reduce complexity of (discovered) models and event data. Imagine a process intelligence project that aims to analyze a model discovered from event data which is further abstracted, possibly multiple times, to reach optimality goals, e.g., reducing model size. So far, after discovering the model, there is no technique that enables the synchronized abstraction of the underlying event log. This results in loosing the grounding in the real-world behavior contained in the log and, in turn, restricts analysis insights. Hence, in this work, we provide the formal basis for synchronized model and event abstraction, i.e., we prove that abstracting a process model by MA and discovering a process model from an abstracted event log yields an equivalent process model. We prove the feasibility of our approach based on behavioral profile abstraction as non-order preserving MA technique, resulting in a novel EA technique.

Figures

Figures reproduced from arXiv: 2505.23536 by the authors.

Figure 1
Figure 1. Process Intelligence in the Financial Domain (Example) [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Running example: Process tree M and abstracted process tree mabpapMq [37] for threshold wt“0.5. ref trace σex1 xRBP,CBW,NC,RP,ATy σex2 xRBP,CBW,NC,RFI,BC,PN,CD,CD,PDD,SC,ATy σex3 xRBP,CBW,NC,RFI,PN,BC,CD,CD,PDD,SC,ATy σex4 xRBP,CBW,NC,PN,RFI,BC,CD,CD,PDD,SC,ATy [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Modular decomposition of two order relation graphs [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

50 extracted references · 50 canonical work pages

  1. [1]

    IEEE TKDE16(9), 1128–1142 (2004)

    van der Aalst, W.M.P., Weijters, T., Maruster, L.: Workflow mining: discovering process models from event logs. IEEE TKDE16(9), 1128–1142 (2004)

    work page 2004

  2. [2]

    The American Mathe- matical Monthly93(5), 333–348 (1986)

    Aldous, D., Diaconis, P.: Shuffling Cards and Stopping Times. The American Mathe- matical Monthly93(5), 333–348 (1986)

    work page 1986

  3. [3]

    In: 2020 International Joint Conference on Neural Networks (IJCNN)

    Angelastro, S., Ferilli, S.: Process Model Modularization by Subprocess Discovery. In: 2020 International Joint Conference on Neural Networks (IJCNN). pp. 1–8. IEEE (2020)

    work page 2020

  4. [4]

    INIS7(23) (4 2020)

    Cicirello, V.A.: Kendall tau sequence distance: Extending kendall tau from ranks to sequences. INIS7(23) (4 2020)

    work page 2020

  5. [5]

    De San Pedro, J., Carmona, J., Cortadella, J.: Log-Based Simplification of Process Models. In: BPM. pp. 457–474. Springer (2015) 40 J.-V. Benzin et al

    work page 2015

  6. [6]

    Fahland, D., van der Aalst, W.M.P.: Simplifying Mined Process Models: An Approach Based on Unfoldings. In: BPM. pp. 362–378. Springer (2011)

    work page 2011

  7. [7]

    Information Systems38(4), 585–605 (2013)

    Fahland,D.,vanderAalst,W.M.P.:Simplifyingdiscoveredprocessmodelsinacontrolled manner. Information Systems38(4), 585–605 (2013)

    work page 2013

  8. [8]

    Ferilli,S.:WoMan:Logic-BasedWorkflowLearningandManagement.IEEETransactions on Systems, Man, and Cybernetics: Systems44(6), 744–756 (2014)

    work page 2014

Show all 50 references
  1. [9]

    In: ATAED

    Janssenswillen, G., Depaire, B., Jouck, T.: Calculating the Number of Unique Paths in a Block-Structured Process Model. In: ATAED. pp. 138–152. CEUR (2016)

    2016

  2. [10]

    In: BPMDS, vol

    Kammerer, K., Kolb, J., Reichert, M.: PQL - A Descriptive Language for Querying, Abstracting and Changing Process Models. In: BPMDS, vol. 214, pp. 135–150. Springer (2015)

    2015

  3. [11]

    In: Service-Oriented Computing

    Kolb, J., Kammerer, K., Reichert, M.: Updatable Process Views for User-Centered Adaption of Large Process Models. In: Service-Oriented Computing. pp. 484–498. Springer (2012)

    2012

  4. [12]

    Kolb, J., Reichert, M.: Data flow abstractions and adaptations through updatable process views. pp. 1447–1453. SAC ’13, ACM (2013)

    2013

  5. [13]

    ACM SIGAPP Applied Computing Review13(1), 6–18 (2013)

    Kolb, J., Reichert, M.: A flexible approach for abstracting and personalizing large business process models. ACM SIGAPP Applied Computing Review13(1), 6–18 (2013)

    2013

  6. [14]

    In: Database and Expert Systems Applications

    Köpke, J., Eder, J., Künstner, M.: Projections of Abstract Interorganizational Business Processes. In: Database and Expert Systems Applications. pp. 472–479. Springer (2014)

    2014

  7. [15]

    440 (2022)

    Leemans, S.J.J.: Robust Process Mining with Guarantees: Process Discovery, Confor- mance Checking and Enhancement, LNBIP, vol. 440 (2022)

    2022

  8. [16]

    Leemans, S.J.J., Fahland, D., van der Aalst, W.M.P.: Discovering Block-Structured Process Models from Event Logs - A Constructive Approach. Tech. rep., Eindhoven University of Technology (2013), no. BPM-13-06

    2013

  9. [17]

    In: Colom, J.M., Desel, J

    Leemans, S.J.J., Fahland, D., van der Aalst, W.M.P.: Discovering Block-Structured Process Models from Event Logs - A Constructive Approach. In: Colom, J.M., Desel, J. (eds.) Application and Theory of Petri Nets and Concurrency. pp. 311–329. LNCS, Springer (2013)

    2013

  10. [18]

    DKE154, 102352 (2024)

    Lim, J., Song, M.: A framework for understanding event abstraction problem solving: Current states of event abstraction studies. DKE154, 102352 (2024)

    2024

  11. [19]

    In: 2024 6th International Conference on Process Mining (ICPM)

    López-Pintado, O., Murashko, S., Dumas, M.: Discovery and Simulation of Data-Aware Business Processes. In: 2024 6th International Conference on Process Mining (ICPM). pp. 105–112 (Oct 2024)

    2024

  12. [20]

    Mafazi, S., Grossmann, G., Mayer, W., Schrefl, M., Stumptner, M.: Consistent Ab- straction of Business Processes Based on Constraints. J. Data Semant.4(1), 59–78 (2015)

    2015

  13. [21]

    In: KiBP@ KR

    Mafazi, S., Mayer, W., Grossmann, G., Stumptner, M.: A Knowledge-based Approach to the Configuration of Business Process Model Abstractions. In: KiBP@ KR. pp. 60–74 (2012)

    2012

  14. [22]

    Discrete Applied Mathematics145(2), 198–209 (Jan 2005)

    McConnell, R.M., de Montgolfier, F.: Linear-time modular decomposition of directed graphs. Discrete Applied Mathematics145(2), 198–209 (Jan 2005)

    2005

  15. [23]

    In: Conceptual Modeling

    Meyer, A., Weske, M.: Data Support in Process Model Abstraction. In: Conceptual Modeling. pp. 292–306. Springer (2012)

    2012

  16. [24]

    Cambridge University Press (2009)

    Milner, R.: The space and motion of communicating agents. Cambridge University Press (2009)

    2009

  17. [25]

    Information Systems114, 102157 (2023)

    Ordoni, E., Mülle, J., Böhm, K.: Reduction of data-value-aware process models: A relevance-based approach. Information Systems114, 102157 (2023)

    2023

  18. [26]

    In: BPSC, pp

    Polyvyanyy, A., Smirnov, S., Weske, M.: On application of structural decomposition for process model abstraction. In: BPSC, pp. 110–122 (2009) Synchronizing Process Model and Event Abstraction 41

    2009

  19. [27]

    A business process archeology tool

    Pérez-Castillo, R., Fernández-Ropero, M., Guzmán, I.G.R.d., Piattini, M.: MARBLE. A business process archeology tool. In: 2011 27th IEEE International Conference on Software Maintenance (ICSM). pp. 578–581 (Sep 2011)

    2011

  20. [28]

    An industrial evaluation

    Pérez-Castillo, R., Fernández-Ropero, M., Piattini, M.: Business process model refactor- ing applying IBUPROFEN. An industrial evaluation. Journal of Systems and Software 147, 86–103 (2019)

    2019

  21. [29]

    In: BPM Workshops, vol

    Reichert, M.: Visualizing Large Business Process Models: Challenges, Techniques, Ap- plications. In: BPM Workshops, vol. 132, pp. 725–736. Springer (2013)

    2013

  22. [30]

    Reichert, M., Kolb, J., Bobrik, R., Bauer, T.: Enabling personalized visualization of large business processes through parameterizable views. pp. 1653–1660. SAC ’12, ACM (2012)

    2012

  23. [31]

    Senderovich, A., Shleyfman, A., Weidlich, M., Gal et al., A.: P^3-Folder: Optimal Model Simplification for Improving Accuracy in Process Performance Prediction. In: BPM. pp. 418–436. Springer (2016)

    2016

  24. [32]

    Information Systems78, 96–111 (Nov 2018)

    Senderovich, A., Shleyfman, A., Weidlich, M., Gal, A., Mandelbaum, A.: To aggregate or to eliminate? Optimal model simplification for improved process performance prediction. Information Systems78, 96–111 (Nov 2018)

    2018

  25. [33]

    Smirnov, S., Dijkman, R., Mendling, J., Weske, M.: Meronymy-Based Aggregation of Activities in Business Process Models. In: ER. pp. 1–14. Springer (2010)

    2010

  26. [34]

    Information Systems37(8), 784–797 (2012)

    Smirnov, S., Reijers, H.A., Weske, M.: From fine-grained to abstract process models: A semantic approach. Information Systems37(8), 784–797 (2012)

    2012

  27. [35]

    Smirnov, S., Reijers, H.A., Weske, M., Nugteren, T.: Business process model abstraction: a definition, catalog, and survey. Distrib. Parallel Databases30(1), 63–99 (2012)

    2012

  28. [36]

    In: Service-Oriented Computing

    Smirnov, S., Weidlich, M., Mendling, J.: Business Process Model Abstraction Based on Behavioral Profiles. In: Service-Oriented Computing. pp. 1–16. Springer (2010)

    2010

  29. [37]

    Smirnov, S., Weidlich, M., Mendling, J.: Business process model abstraction based on synthesis from well-structured behavioral profiles. Int. J. Coop. Inf. Syst.21(01), 55–83 (2012)

    2012

  30. [38]

    In: Proceedings of the 19th Panhellenic Conference on Informatics

    Tsagkani, C., Tsalgatidou, A.: Abstracting BPMN models. In: Proceedings of the 19th Panhellenic Conference on Informatics. pp. 243–244. PCI ’15, Association for Computing Machinery (2015)

    2015

  31. [39]

    Information Systems103, 101818 (Jan 2022)

    Tsagkani, C., Tsalgatidou, A.: Process model abstraction for rapid comprehension of complex business processes. Information Systems103, 101818 (Jan 2022)

    2022

  32. [40]

    In: BMSD

    Tsalgatidou, A., Tsagkani, C.: Rule-based Business Process Abstraction Framework:. In: BMSD. pp. 173–178. SCITEPRESS (2016)

    2016

  33. [41]

    Van Houdt, G., de Leoni, M., Martin, N., Depaire, B.: An empirical evaluation of unsupervised event log abstraction techniques in process mining. Inf. Syst.121, 102320 (2024)

    2024

  34. [42]

    Data & Knowledge Engineering68(9), 793–818 (Sep 2009)

    Vanhatalo, J., Völzer, H., Koehler, J.: The refined process structure tree. Data & Knowledge Engineering68(9), 793–818 (Sep 2009)

    2009

  35. [43]

    Völker, M., Weske, M.: Ontology-Based Abstraction of Bot Models in Robotic Process Automation. In: ER. pp. 239–256 (2023)

    2023

  36. [44]

    International Journal of Cooperative Information Systems 28(03), 1950007 (2019)

    Wang, N., Sun, S., Liu, Y., Zhang, S.: Business Process Model Abstraction Based on Fuzzy Clustering Analysis. International Journal of Cooperative Information Systems 28(03), 1950007 (2019)

    2019

  37. [45]

    Business & Information Systems Engineering 60(6), 525–542 (2018)

    Wang, N., Sun, S., OuYang, D.: Business Process Modeling Abstraction Based on Semi-Supervised Clustering Analysis. Business & Information Systems Engineering 60(6), 525–542 (2018)

    2018

  38. [46]

    Computers in Industry62(5), 467–486 (2011) 42 J.-V

    Weber, B., Reichert, M., Mendling, J., Reijers, H.A.: Refactoring large process model repositories. Computers in Industry62(5), 467–486 (2011) 42 J.-V. Benzin et al

    2011

  39. [47]

    In: PETRI NETS

    van der Werf, J.M.E.M., van Dongen, B.F., Hurkens, C.A.J., Serebrenik, A.: Process Discovery Using Integer Linear Programming. In: PETRI NETS. pp. 368–387 (2008)

    2008

  40. [48]

    In: ICCIT

    Xue, G., Zhang, K., Yang, J., Yao, S.: Plain abstraction of business process model. In: ICCIT. pp. 338–341 (2011)

    2011

  41. [49]

    Ye, J., Zhang, S., Lin, Y.: Log Optimization Simplification Method for Predicting Remaining Time (2025), arXiv:2503.07683 [cs]

    2025 arXiv

  42. [50]

    van Zelst, S.J., Mannhardt, F., de Leoni, M., Koschmider, A.: Event abstraction in process mining: literature review and taxonomy. Gran. Comp.6(3), 719–736 (2021)

    2021

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.