Bpm10slides
- 1. A fresh look at Precision in Process ConformanceJorge Muñoz-GamaJosep CarmonaUniversitatPolitècnica de Catalunya(Barcelona, Spain)
- 2. Outline15 Sep 2010Precision in Process Conformance2Process Mining and Process ConformanceMotivationApproachGeneral ApproachImplementationResultsExtensionsFuture workConclusions
- 3. Process Mining15 Sep 2010Precision in Process Conformance3* www.processmining.org
- 4. Conformance Dimensions15 Sep 2010Precision in Process Conformance4How much of the observed behavior is captured by the modelModels with minimal behavior to represent accurately the logOverly precise models which overfit the logMinimal structure which clearly reflect the behavior
- 5. Outline15 Sep 2010Precision in Process Conformance5Process Mining and Process ConformanceMotivationApproachGeneral ApproachImplementationResultsExtensionsFuture workConclusions
- 6. Related Work15 Sep 2010Precision in Process Conformance6Precision in the literatureMost related work Rozinat et al. Information System 33 (2008)Metric for Precision in Petri NetsComputation of Follows and Precedes relations (Always, Never, Sometimes) of Model and Log.Measurement based on discrepancies in Sometimes relationsModel relations require a model space state explorationCoverabilityGraph
- 7. Motivation15 Sep 2010Precision in Process Conformance7Goals and RequirementsPrecision DimensionPetri NetsAvoid the complete state space explorationEffort needed to obtain an accurate modelFine-level precisionLocate the precision inconsistencies
- 8. Process Conformance and Refinement15 Sep 2010Precision in Process Conformance8Locate the inconsistenciesPetri NetConformance(Precision)BADCMDTETC PrecisionMetricA B DA C DMeasure the inconsistenciesEvent Log
- 9. Outline15 Sep 2010Precision in Process Conformance9Process Mining and Process ConformanceMotivationApproachGeneral ApproachImplementationResultsExtensionsFuture workConclusions
- 10. General Idea: Escaping Edges15 Sep 2010Precision in Process Conformance10ModelBehaviorEscapingEdgesLogBehaviorModelBehavior
- 11. Conformance Route Map15 Sep 2010Precision in Process Conformance11Petri NetBADCMDTModelStatesTraversalMetricLogStatesA B DA C DEvent Log
- 12. Log and Model States15 Sep 2010Precision in Process Conformance12Log Incorporate state information in the log(Aalst et al. Software and Systems Modeling, 2009)Past, Unlimited and SequenceModelMarkings of the Petri Net
- 13. Model States and Mapping15 Sep 2010Precision in Process Conformance13Not all the reachable markings (could be infinite)Only Markings with a Log State mapped onLog and Model States Mappingi.e., reached marking after replay state prefix p2p3p4p1p1 p2 p3 p4 p5BA B Es2s1s3s40 1 0 0 1…0 1 0 0 nEACp5DMarkings not exploredp1p4p3p2
- 14. Log-guided Traversal15 Sep 2010Precision in Process Conformance14Log-guided Traversal of Model BehaviorAllowed Tasks : i.e., actions enabled in that momentReflected Tasks : i.e., actions really executed (thus, annotated in the log) BCD<p2>p2p3p4p1A B EA C EBBBCp2p3p4p1EEACACA B EA C EDD
- 15. Traversal (2)15 Sep 2010Precision in Process Conformance15Escaping Edges : i.e., enabled actions not executedPrecision discrepanciesBCDBp1p2p3p4EACBCDA B EA C ED
- 16. Precision Metric15 Sep 2010Precision in Process Conformance16Take into account the Escaping EdgesBetween 0 (imprecise) and 1 (precise)More frequent traces, more weight in the metricIndependent of Structural dimensionGlobal precisionLocalizabilityA H I ZA P Q ZHIAZPQ
- 17. Minimal Disconformant Traces (MDT)15 Sep 2010Precision in Process Conformance17Localizability of precision inconsistenciesi.e., Minimal traces indicating where the model starts to deviate from the logAlgorithm to compute all MDT using Escaping Edges BDACMDTA EA B EC D P QRefined Petri Net
- 18. Outline15 Sep 2010Precision in Process Conformance18Process Mining and Process ConformanceMotivationApproachGeneral ApproachImplementationResultsExtensionsFuture workConclusions
- 19. Implementation15 Sep 2010Precision in Process Conformance19ProM 6 FrameworkETConformance Plug-In
- 20. Outline15 Sep 2010Precision in Process Conformance20Process Mining and Process ConformanceMotivationApproachGeneral ApproachImplementationResultsExtensionsFuture workConclusions
- 21. Results15 Sep 2010Precision in Process Conformance21
- 22. Results (2)15 Sep 2010Precision in Process Conformance22
- 23. Outline15 Sep 2010Precision in Process Conformance23Process Mining and Process ConformanceMotivationApproachGeneral ApproachImplementationResultsExtensionsFuture workConclusionsInvisible Tasks
- 24. Duplicate Tasks
- 26. Non fittingdonedonein progressin progress
- 27. Invisible Tasks15 Sep 2010Precision in Process Conformance24(Transitions associated with no event)p3IBWhich Sequence?
- 28. A H C ?
- 29. A I C?AHCp4INDETERMINISM A C
- 30. Invisible Tasks (2)15 Sep 2010Precision in Process Conformance25Invisible Coverability GraphSolutionsUnion of EnabledLazy Invisibles *One path onlyShortest Invisible Path *A,BAB<1, 0, 0>Inv2Inv1<1, ω, 0><0, 0, 1>CCInv3A,DDX<0, ω, 1>A,CXX*Rozinat et al.Information System 33 (2008)
- 31. Duplicate Tasks15 Sep 2010Precision in Process Conformance26(Several Transitions associated with the same event)Which Task?B ?B ?INDETERMINISMSolutionse.g. Look-aheadBCABD... A B C ...
- 32. Variant: States as Markings15 Sep 2010Precision in Process Conformance27States as Prefix2Escaping EdgesBCBACA B CStates as MarkingsBA B CACNO Escaping Edgesp1p2p3<p1><p2><p3>
- 33. Variant: Non fitting models15 Sep 2010Precision in Process Conformance28Symmetric to the Escaping Edges (Ee)Log Escaping Edges (LEe): The points where the log deviates from the modelFitness instead of PrecisionModelBehaviorEscapingEdgesLogBehaviorLogEscapingEdgesModelBehavior
- 34. Outline15 Sep 2010Precision in Process Conformance29Process Mining and Process ConformanceMotivationApproachGeneral ApproachImplementationResultsExtensionsFuture workConclusions
- 35. Future Work: Refinement15 Sep 2010Precision in Process Conformance30BADCMDTA EA B EB H J GRefined Petri NetEvent LogBA EA B EDAPetri NetC
- 36. Concurrencies in the model but not in the logBreak the model concurrency with a restriction, e.g. a placeStructural ConcurrencyBest effort overapproximation for general Petri NetsExact for live and bounded Free Choice systemsPolynomial Algorithm Kovalyov and Esparza , Proc. Intl. Workshop on Discrete Event Sytems, 1996Future Work: Breaking Concurrencies15 Sep 2010Precision in Process Conformance31BDAA B C DC
- 37. Supervisory Control15 Sep 2010Precision in Process Conformance32Refined ModelMDT AbstractionMDTSupervisorModelSupervisory Control in Process MiningSantos et al. Supervisory Control Service (2010)
- 38. Conclusions15 Sep 2010Precision in Process Conformance33New technique for precision between Petri nets and Log.Avoids complete models state space exploration.Metric based on the effort needed to obtain a precise model.MDT, indicating the points where the model starts to deviates from the log.Approach implemented as plug-in of ProM 6.
- 39. Thank You15 Sep 2010Precision in Process Conformance34Thank You for Your Attention