Distributed Process Discovery and Conformance Checking
1 like823 views
The document discusses process discovery and conformance checking. It begins with an introduction to different roles of process models and examples of process discovery on real event logs. It then covers topics like replay, conformance checking, and analyzing models based on criteria like fitness and simplicity. Process discovery algorithms discussed include state-based regions and language-based regions approaches. The document explains how conformance checking involves replaying traces and calculating alignments between event logs and models.
![Example: Process Discovery Using
State-Based Regions
d
e
[a,e] [a,d,e]
[ a,b]
a b
[] [a] c
c
b d
[a,c] [a,b,c] [a,b,c,d]
b
a p1 e p3 d
start end
p2 c p4
PAGE 33](https://image.slidesharecdn.com/fase-2012-invited-talk-wvda-120327090511-phpapp01/85/Distributed-Process-Discovery-and-Conformance-Checking-34-320.jpg)
![Example of Region
d
e
[a,e] [a,d,e]
[ a,b]
a b
[] [a] c
c
b d
[a,c] [a,b,c] [a,b,c,d]
enter: b,e
leave: d
do-not-cross: a,c
b
a p1 e p3 d
start end
p2 c p4
PAGE 34](https://image.slidesharecdn.com/fase-2012-invited-talk-wvda-120327090511-phpapp01/85/Distributed-Process-Discovery-and-Conformance-Checking-35-320.jpg)
Distributed Process Discovery and Conformance Checking
- 1. Distributed Process Discovery and Conformance Checking prof.dr.ir. Wil van der Aalst www.processmining.org
- 2. On the different roles of (process) models … PAGE 1
- 3. Play-Out PAGE 2
- 4. Play-Out (Classical use of models) A B C D AED AED ABCD ACBD ACBD AED ACBD PAGE 3
- 5. Play-In PAGE 4
- 6. Play-In ABCD AED AED ABCD ACBD ACBD AED ACBD PAGE 5
- 7. Example Process Discovery (Vestia, Dutch housing agency, 208 cases, 5987 events) PAGE 6
- 8. Example Process Discovery (ASML, test process lithography systems, 154966 events) PAGE 7
- 9. Example Process Discovery (AMC, 627 gynecological oncology patients, 24331 events) PAGE 8
- 10. Replay PAGE 9
- 11. Replay ABCD PAGE 10
- 12. Replay AED PAGE 11
- 13. Replay can detect problems AC D Problem! Problem! token left behind missing token PAGE 12
- 14. Conformance Checking (WOZ objections Dutch municipality, 745 objections, 9583 event, f= 0.988) PAGE 13
- 15. Replay can extract timing information A5B8 C9 D13 8 5 6 4 7 3 2 5 8 5 13 4 3 4 37 4 7 6 9 PAGE 14
- 16. Performance Analysis Using Replay (WOZ objections Dutch municipality, 745 objections, 9583 event, f= 0.988) PAGE 15
- 17. Big Data PAGE 16
- 18. “All of the world's Big Data music can be stored on a $600 disk drive.” “Enterprises globally stored more than 7 exabytes of new data on disk drives in 2010, while consumers stored more than 6 exabytes of new data on “Indeed, we are devices such as generating so much PCs and data today that it is notebooks.” physically impossible to store it all. Health care providers, for instance, discard 90 percent of the data that they generate.” Source: “Big Data: The Next Frontier for Innovation, Competition, and Productivity” McKinsey Global Institute, 2011. PAGE 17
- 19. Hilbert and Lopez. The World's Technological Capacity to Store, Communicate, and Compute Information. Science, 332(6025):60-65, 2011. PAGE 18
- 20. www.olifantenpaadjes.nl PAGE 19
- 21. PAGE 20
- 22. PAGE 21
- 23. Evidence-Based Computer Science PAGE 22
- 24. How good is my model? PAGE 23
- 25. Four Competing Quality Criteria “able to replay event log” “Occam’s razor” “not overfitting the log” “not underfitting the log” PAGE 24
- 26. Example: one log four models b examine thoroughly g pay c compensation a examine e start register casually decide end # trace request h 455 acdeh d reject check ticket request 191 abdeg f reinitiate request 177 adceh N1 : fitness = +, precision = +, generalization = +, simplicity = + 144 abdeh 111 acdeg a c d e h 82 adceg start register examine check decide reject end request casually ticket request 56 adbeh N2 : fitness = -, precision = +, generalization = -, simplicity = + 47 acdefdbeh “able to replay event log” “Occam’s razor” 38 adbeg examine check thoroughly b d ticket g 33 acdefbdeh fitness simplicity pay compensation a 14 acdefbdeg start register examine c end 11 acdefdbeg request casually e f reinitiate h process decide request reject request 9 adcefcdeh discovery N3 : fitness = +, precision = -, generalization = +, simplicity = + 8 adcefdbeh 5 adcefbdeg a d c e g 3 acdefbdefdbeg generalization precision register request check ticket examine casually decide pay compensation 2 adcefdbeg a c d e g 2 adcefbdefbdeg “not overfitting the log” “not underfitting the log” register examine check decide pay request casually ticket compensation 1 adcefdbefbdeh a d c e h 1 adbefbdefdbeg register check examine decide reject request ticket casually request 1 adcefdbefcdefdbeg a c d e h 1391 start end register examine check decide reject request casually ticket request (all 21 variants seen in the log) a b d e g register examine check decide pay request thoroughly ticket compensation a d b e h register check examine decide reject request ticket thoroughly request a b d e h register examine check decide reject request thoroughly ticket request PAGE 25 N4 : fitness = +, precision = +, generalization = -, simplicity = -
- 27. # trace 455 acdeh Model N1 191 abdeg 177 adceh 144 abdeh 111 acdeg 82 adceg 56 adbeh 47 acdefdbeh 38 adbeg 33 acdefbdeh 14 acdefbdeg 11 acdefdbeg 9 adcefcdeh 8 adcefdbeh 5 adcefbdeg 3 acdefbdefdbeg 2 adcefdbeg 2 adcefbdefbdeg 1 adcefdbefbdeh 1 adbefbdefdbeg 1 adcefdbefcdefdbeg PAGE 26 1391
- 28. # trace 455 acdeh Model N2 191 abdeg 177 adceh 144 abdeh 111 acdeg 82 adceg 56 adbeh 47 acdefdbeh 38 adbeg 33 acdefbdeh 14 acdefbdeg 11 acdefdbeg 9 adcefcdeh 8 adcefdbeh 5 adcefbdeg 3 acdefbdefdbeg 2 adcefdbeg 2 adcefbdefbdeg 1 adcefdbefbdeh 1 adbefbdefdbeg 1 adcefdbefcdefdbeg PAGE 27 1391
- 29. # trace 455 acdeh Model N3 191 abdeg 177 adceh 144 abdeh 111 acdeg 82 adceg 56 adbeh 47 acdefdbeh 38 adbeg 33 acdefbdeh 14 acdefbdeg 11 acdefdbeg 9 adcefcdeh 8 adcefdbeh 5 adcefbdeg 3 acdefbdefdbeg 2 adcefdbeg 2 adcefbdefbdeg 1 adcefdbefbdeh 1 adbefbdefdbeg 1 adcefdbefcdefdbeg PAGE 28 1391
- 30. # trace 455 acdeh Model N4 191 abdeg 177 adceh 144 abdeh a d c e g 111 acdeg register check examine decide pay request ticket casually compensation 82 adceg a c d e g 56 adbeh register examine check decide pay request casually ticket compensation 47 acdefdbeh a d c e h 38 adbeg register check examine decide reject request ticket casually request 33 acdefbdeh a c d e h 14 acdefbdeg start end register examine check decide reject request casually ticket request 11 acdefdbeg 9 adcefcdeh (all 21 variants seen in the log) 8 adcefdbeh 5 adcefbdeg a b d e g register examine check decide pay 3 acdefbdefdbeg request thoroughly ticket compensation 2 adcefdbeg a d b e h register check examine decide reject 2 adcefbdefbdeg request ticket thoroughly request 1 adcefdbefbdeh a b d e h register examine check decide reject 1 adbefbdefdbeg request thoroughly ticket request 1 adcefdbefcdefdbeg N4 : fitness = +, precision = +, generalization = -, simplicity = - PAGE 29 1391
- 31. Process Discovery PAGE 30
- 32. Process Discovery (small selection) distributed genetic mining automata-based learning language-based regions heuristic mining genetic mining state-based regions LTL mining stochastic task graphs neural networks fuzzy mining hidden Markov models mining block structures α algorithm conformal process graph multi-phase mining partial-order based mining α# algorithm ILP mining α++ algorithm PAGE 31
- 33. Petri net view: Just discover the places … Adding a place limits behavior: • overfitting ≈ adding too many places • underfitting ≈ adding too few places PAGE 32
- 34. Example: Process Discovery Using State-Based Regions d e [a,e] [a,d,e] [ a,b] a b [] [a] c c b d [a,c] [a,b,c] [a,b,c,d] b a p1 e p3 d start end p2 c p4 PAGE 33
- 35. Example of Region d e [a,e] [a,d,e] [ a,b] a b [] [a] c c b d [a,c] [a,b,c] [a,b,c,d] enter: b,e leave: d do-not-cross: a,c b a p1 e p3 d start end p2 c p4 PAGE 34
- 36. Example: Process Discovery Using Language-Based Regions A place is feasible if it can be added without disabling any of the traces in the event log. R PAGE 35
- 37. Conformance Checking PAGE 36
- 38. Replaying trace “abeg” a b e g r=1 m=1 1 1 = 0.83333 6 6 PAGE 37
- 39. # trace 455 acdeh Can be lifted to log level 191 abdeg 177 adceh N1 b 144 abdeh examine thoroughly g 111 acdeg p1 pay c p3 compensation 82 adceg a examine e start register casually decide p5 end 56 adbeh request h p2 d p4 reject 47 acdefdbeh check ticket request f reinitiate 38 adbeg request 33 acdefbdeh 14 acdefbdeg 11 acdefdbeg 9 adcefcdeh 8 adcefdbeh 5 adcefbdeg 3 acdefbdefdbeg 2 adcefdbeg 2 adcefbdefbdeg 1 adcefdbefbdeh 1 adbefbdefdbeg 1 adcefdbefcdefdbeg PAGE 38 1391
- 40. From “playing the token game” to optimal alignments … observed trace: “abeg” a b » e g a b d e g move in model only PAGE 39
- 41. Another alignment observed trace: “abcdeg” a b c d e g a b » d e g move in log only PAGE 40
- 42. Moves in an alignment move in log trace in event log a b » d e g a » c d e g possible run of model move in model move in both Optimal alignment describes modeled behavior closest to observed behavior PAGE 41
- 43. Moves have costs … a … … » … … » … … a … … a … … a … … a … … b … • Standard cost function: − c(x,») = 1 − c(»,y) = 1 − c(x,y) = 0, if x=y − c(x,y) = ∞, if x≠y PAGE 42
- 44. Non-fitting trace: abefdeg abefdeg a b » e f d » e g 2 a b d e f d b e g a b e f d e g 2 a b » » d e g PAGE 43
- 45. Any cost structure is possible … send-letter(John,2 … weeks, $400) … send-email(Sue,3 … weeks,$500) • Similar activities (more similarity implies lower costs). • Resource conformance (done by someone that does not have the specified role). • Data conformance (path is not possible for this customer). • Time conformance (missed the legal deadline) PAGE 44
- 46. b examine thoroughly g pay c compensation Fitness a e 1.0 examine start register casually decide end # trace request h 455 acdeh d reject check ticket request 191 abdeg f reinitiate request 177 adceh N1 : fitness = +, precision = +, generalization = +, simplicity = + 144 abdeh 111 acdeg a c d e h 82 adceg Our A* algorithm 0.8 start register request examine casually check ticket N2 : fitness = -, precision = +, generalization = -, simplicity = + decide reject request end 56 adbeh exploits the Petri 47 acdefdbeh 38 adbeg net marking examine thoroughly b d check ticket pay g 33 acdefbdeh equation and uses a compensation 14 acdefbdeg other “tricks” to 1.0 start register request examine casually c decide e f reinitiate request reject request h end 11 acdefdbeg 9 adcefcdeh prune the search N3 : fitness = +, precision = -, generalization = +, simplicity = + 8 adcefdbeh 5 adcefbdeg space. a d c e g 3 acdefbdefdbeg register check examine decide pay request ticket casually compensation 2 adcefdbeg a c d e g 2 adcefbdefbdeg register examine check decide pay request casually ticket compensation 1 adcefdbefbdeh a d c e h 1 adbefbdefdbeg register check examine decide reject request ticket casually request 1 adcefdbefcdefdbeg 1.0 start a register request c examine casually d check ticket e decide h reject request end 1391 (all 21 variants seen in the log) a b d e g register examine check decide pay Aligned event log is request thoroughly ticket compensation a d b e h starting point for other register request check ticket examine thoroughly decide reject request types of analysis. a register b d check e decide h reject examine request thoroughly ticket request PAGE 45 N4 : fitness = +, precision = +, generalization = -, simplicity = -
- 47. Distributing/Decomposing “Big Data” Process Mining Problems PAGE 46
- 48. PAGE 47
- 49. What if? there are more than 100.000.000 events? there are more than 1000 different activities? acefgijkl conformance add extra acddefhkjil checking insurance g abdefjkgil c8 process c4 acdddefkhijl discovery acefgijkl abefgjikl h skip extra ... b insurance skip extra change c5 c9 insurance d booking i a c select car in book car c1 add extra c2 c6 insurance e f j l confirm c3 check driver’s c10 out initiate supply check-in license car there are more k than 1.000.000 c7 charge credit c11 cases? card PAGE 48
- 50. Distributed computing • multicore CPU • manycore GPU • cluster computing • grid computing • cloud computing • … PAGE 49
- 51. How to distribute process discovery? PAGE 50
- 52. How to distribute conformance checking? f abcdeg adcefbcfdeg abdceg abcdefbcdeg c abdfcefdceg acdefbdceg a b c2 c4 e g abcdeg abdceg in c1 d c6 out abdcefbdcefbdceg abcdeg c3 c5 abcdefbcdefbdceg abcdefbdceg acdefg adcfeg abdcefcdfeg abcdeg abcdeg abdceg abcdefbcdeg f occurs abcdeg too often abdceg abdcefbdcefbdceg f abcdeg abcdefbcdefbdceg abcdefbdceg abcdeg c a b c2 c4 e g in c1 d c6 out c3 c5 adcefbcfdeg abdfcefdceg b is often acdefbdceg skipped acdefg adcfeg PAGE 51 abdcefcdfeg
- 53. Classification based on partitioning of event log: vertical and horizontal sets of cases sets of activities PAGE 52
- 54. Replication: Same event log on all computing nodes Only makes sense if random elements, e.g., genetic process mining. PAGE 53
- 55. Vertical distribution I: Split cases arbitrarily sets of cases abcdeg abdcefbdcefbdceg abcdeg abdcefbcdeg abcdeg abdcefbcdeg abdceg abcdefbcdefbdceg abdceg abcdefbcdeg abcdefbdceg abcdefbcdeg abdcefbdceg abcdeg abdcefbdceg abcdefbdceg abdceg abcdefbdceg abcdeg abdcefbcdeg abcdeg abdceg abcdeg abdceg abdcefbdcefbdceg abcdeg abcdefbcdefbdceg abcdefbdceg abcdeg abdceg abdcefbcdeg abcdeg PAGE 54
- 56. Vertical distribution II: Split cases based on a specific feature abcdeg abcdeg abdcefbcdeg abdcefbcdeg abdceg abcdefbcdeg abdceg abcdeg abdcefbdceg abcdefbcdeg abdceg abcdefbdceg abdcefbdceg abcdeg abcdefbdceg abcdefbdceg abcdeg abdcefbcdeg abcdeg abdceg abdceg abcdeg abdcefbdcefbdceg abcdeg abcdefbcdefbdceg abcdefbdceg abcdeg abdceg abdcefbdcefbdceg abdcefbcdeg abcdefbcdefbdceg abcdeg PAGE 55
- 57. Horizontal distribution sets of activities PAGE 56
- 58. Horizontal distribution: The key idea projected on projected on a,b,e,f,g b,c,d,e PAGE 57
- 59. Passages for Horizontal Distribution PAGE 58
- 60. Passages PAGE 59
- 61. causal dependency: may trigger or enable Passage P=(X,Y) PAGE 60
- 62. Minimal passages a passage is minimal if it does not contain smaller passages PAGE 61
- 63. Passages define an equivalence relation on the edges in the graph PAGE 62
- 64. Minimal passage 1: ({a},{b,c}) PAGE 63
- 65. Minimal passage 2: ({b,c,d},{d,e,f}) PAGE 64
- 66. Minimal passage 3: ({e},{g}) PAGE 65
- 67. Minimal passage 4: ({f},{h}) PAGE 66
- 68. Minimal passage 5: ({g,h},{i}) PAGE 67
- 69. So What? • Any process model can be partitioned in minimal passages. • Discovery and conformance checking can be done per passage! clouds may contain a d f h arbitrary subprocesses not k n explicitly recorded in the event log (invisible activities o or small networks used for routing, e.g. XOR/AND/OR- b e i split/joins) l i g p o c j m PAGE 68
- 70. Example result for Petri nets f a d h k n “The event log fits all o passages if and only if b e i the event log fits the i g l p o whole model.” c j m Key insight: interface transitions controlled by event log PAGE 69
- 71. Discovery example a g in out f f b e a g c c causal structure obtained using b e heuristics & domain knowledge d d f c a b c2 c4 e g in c1 d c6 out c3 c5 PAGE 70
- 72. Conformance checking acefl add extra acddefl insurance g abdefl c4 c8 acdddefl acefl h skip extra abefl b insurance ... skip extra change c5 c9 insurance d booking i a c select car in book car c1 add extra c2 c6 insurance e f j l confirm c3 check driver’s c10 out initiate supply check-in license car k c7 c11 charge credit card PAGE 71
- 73. Create Skeleton PAGE 72
- 74. Net fragments per passage PAGE 73
- 75. Initial implementation in ProM PAGE 74
- 76. Super linear speedups possible (even when using a single computer decomposition helps) PAGE 75
- 77. Conclusion PAGE 76
- 78. Conclusion f a d h k n o b e i l i g p o c j m “Big Data” PAGE 77
- 79. www.processmining.org www.win.tue.nl/ieeetfpm/ PAGE 78