SLIDE 1
33
PAGE 0
Processes @ your Service
Using Process Mining to Turn Big Data into Real Value
prof.dr.ir. Wil van der Aalst
Web Engineering
PAGE 1
model, specify, configure, implement processes
- bserve
behavior (e.g., event data)
Processes @ your Service Using Process Mining to Turn Big Data into - - PowerPoint PPT Presentation
Processes @ your Service Using Process Mining to Turn Big Data into - - PowerPoint PPT Presentation
33 Processes @ your Service Using Process Mining to Turn Big Data into Real Value prof.dr.ir. Wil van der Aalst PAGE 0 Web Engineering model, specify, observe configure, implement behavior (e.g., processes event data) PAGE 1 PAGE 2
SLIDE 2
SLIDE 3
PAGE 2
SLIDE 4
PAGE 3
Big Data ?
SLIDE 5
PAGE 4
Big … or fast and efficient?
SLIDE 6
PAGE 5
The future is bright, but how to get started? What are the main pitfalls of process modeling? What is process mining? Why is process discovery difficult? What are the main research challenges?
SLIDE 7
PAGE 6
The future is bright, but how to get started? What are the main pitfalls of process modeling? What is process mining? Why is process discovery difficult? What are the main research challenges?
SLIDE 8 register request examine casually examine thoroughly check ticket decide pay compensation reject request reinitiate request start end
start register request examine thoroughly examine casually check ticket decide pay compensation reject request end e1
AND OR XOR OR AND XOR
end e2 e3 e4 e5 e6
start register request examine thoroughly examine casually check ticket decide pay compensation reject request new information end c1 c2 OR-split OR-join c3
a
start register request
b
examine thoroughly
c
examine casually
d
check ticket decide pay compensation reject request reinitiate request
e g h f
end c1 c2 c3 c4 c5
[start] register request examine thoroughly examine casually check ticket decide pay compensation reject request reinitiate request [end] [c1,c2] [c1,c4] [c2,c3] [c3,c4] check ticket examine casually examine thoroughly [c5] SLIDE 9
PAGE 8
- enormous investments in process models
- large collections of "dead" process models
- not taken seriously, unrelated to reality
SLIDE 10
PAGE 9
problem #1 Aiming for one model that suits all purposes
SLIDE 11
PAGE 10
SLIDE 12
PAGE 11
problem #2 Straightjacketing smaller interacting processes into one monolithic model
SLIDE 13
PAGE 12
register request examine casually examine thoroughly check ticket decide pay compensation reject request reinitiate request start end
SLIDE 14
What is the process instance?
PAGE 13 Order Customer : CustID Amount : Euro Created : DateTime Paid : DateTime Completed : DateTime Orderline Product : ProdID NofItems : PosInt TotalWeight : Weight Entered : DateTime BackOrdered : DateTime Secured : DateTime Delivery DelAddress : Address Contact : PhoneNo Attempt When : DateTime Successful : Bool
1 1..* 0..* 1 0..1 1..*
OrderID : OrderID OrderID : OrderID OrderLineID : OrderLineID DelID : DelID DelID : DelID DelID : DelID
SLIDE 15
PAGE 14
problem #3 Using a static hierarchical decomposition as the only abstraction mechanism
SLIDE 16
PAGE 15
most process modeling notations assume a fixed hierarchy no seamless zoom-in and zoom out! traditional hierarchy concepts don't support "Google Maps" abstraction
SLIDE 17
PAGE 16
problem #4 Modeling humans as if they are machines doing a single task
SLIDE 18
PAGE 17
"My processes are unique, my people are artists!"
SLIDE 19
PAGE 18
?
SLIDE 20
PAGE 19
problem #5 Being vague about vagueness
SLIDE 21
PAGE 20
register request examine casually examine thoroughly check ticket decide pay compensation reject request reinitiate request start end
SLIDE 22
PAGE 21
problem #6 Abstracting from the things that matter
SLIDE 23
PAGE 22
SLIDE 24
PAGE 23
The future is bright, but how to get started? What are the main pitfalls of process modeling? What is process mining? Why is process discovery difficult? What are the main research challenges?
SLIDE 25
Positioning Process Mining
PAGE 24
process mining
data-oriented analysis
(data mining, machine learning, business intelligence)
process model analysis
(simulation, verification, optimization, gaming, etc.)
performance-
- riented
questions, problems and solutions compliance-
- riented
questions, problems and solutions
SLIDE 26
PAGE 25
www.olifantenpaadjes.nl
SLIDE 27
PAGE 26
SLIDE 28
PAGE 27
Let us take a step back and see how models and behavior relate: Let's play!
SLIDE 29
Play-Out
PAGE 28
event log process model
SLIDE 30
A B C D E p2 end p4 p3 p1 start
Play-Out (Classical use of models)
PAGE 29
A B C D A C B D A B C D A E D A C B D A C B D A E D A E D
SLIDE 31
Play-In
PAGE 30
event log process model
SLIDE 32
A B C D E p2 end p4 p3 p1 start
Play-In
PAGE 31
A C B D A B C D A E D A C B D A C B D A E D A E D A B C D
SLIDE 33
Example Process Discovery
(Vestia, Dutch housing agency, 208 cases, 5987 events)
PAGE 32
SLIDE 34
Example Process Discovery
(ASML, test process lithography systems, 154966 events)
PAGE 33
SLIDE 35
Example Process Discovery
(AMC, 627 gynecological oncology patients, 24331 events)
PAGE 34
SLIDE 36
Replay
PAGE 35
event log process model · extended model showing times, frequencies, etc. · diagnostics · predictions · recommendations
SLIDE 37
A B C D E p2 end p4 p3 p1 start
Replay
PAGE 36
A B C D
SLIDE 38
A B C D E p2 end p4 p3 p1 start
Replay
PAGE 37
A E D
SLIDE 39
A B C D E p2 end p4 p3 p1 start
Replay can detect problems
PAGE 38
A C D
Problem! missing token Problem! token left behind
SLIDE 40
Conformance Checking
(WOZ objections Dutch municipality, 745 objections, 9583 event, f= 0.988)
PAGE 39
SLIDE 41
A B C D E p2 end p4 p3 p1 start
Replay can extract timing information
PAGE 40
A5 B8 C9 D13
5 8 9 13
3 4 5 4
3 2 6 5 8 7 6 4 7 7 4 3
SLIDE 42
PAGE 41
Performance Analysis Using Replay
(WOZ objections Dutch municipality, 745 objections, 9583 event, f= 0.988)
SLIDE 43
PAGE 42
Models are like the glasses required to see and understand event data!
SLIDE 44
PAGE 43
SLIDE 45
PAGE 44
- conformance checking to diagnose deviations
- squeezing reality into the model to do model-based
analysis
Alignments are essential!
SLIDE 46
PAGE 45
process model event log synchronous move move on model only move on log
- nly
SLIDE 47
Example: BPI Challenge 2012
(Dutch financial institute, doi:10.4121/uuid:3926db30-f712-4394-aebc-75976070e91f)
PAGE 46
“O_DECLINED” and “W_Wijzigen contractgegevens” are often skipped Many moves on log of “O_CANCELLED”, ”O_CREATED”, ”O_SELECTED”, “O_SENT” occurred with the same frequency value (i.e. 60) before parallel branch Many moves on log of “W_Afhandelen leads” ( > 2200 times)
- ccurred in the end of
traces Loops of “W_Completeren aanvraag” and “W_Nabellen offertes” are often performed
Work of Arya Adriansyah (Replay project)
SLIDE 48
PAGE 47
“O_DECLINED” and “W_Wijzigen contractgegevens” are often skipped Many moves on log of “O_CANCELLED”, ”O_CREATED”, ”O_SELECTED”, “O_SENT” occurred with the same frequency value (i.e. 60) before parallel branch Many moves on log of “W_Afhandelen leads” ( > 2200 times)
- ccurred in the end of
traces Loops of “W_Completeren aanvraag” and “W_Nabellen offertes” are often performed
Synchronous moves of “Completeren aanvraag” Move on log of “Completeren aanvraag” Moves on model towards end of traces Move on log of “O_CANCELLED” and “A_CANCELLED”
“O_ACCEPTED” has average sojourn time of 27.07 minutes, while “A_REGISTERED”, ”A_ACTIVATED”, and “A_APPROVED” have average sojourn time of 29.56 minutes Activity “W_Wijzigen contractgegevens” is the bottleneck, but it occured rarely (only 4 times) The average waiting time for the input place of “W_Nabellen offertes+START” is very long (2.83 days) compares to the average waiting time of other places
SLIDE 49
PAGE 48
The future is bright, but how to get started? What are the main pitfalls of process modeling? What is process mining? Why is process discovery difficult? What are the main research challenges?
SLIDE 50
Language identification in the limit (Mark Gold 1967)
PAGE 49
Language identification in the limit by E Mark Gold, Information and Control, 10(5):447–474, 1967.
abc abd abc ? ab(c|d) ? ad abbc ac … (ad)|(ab(c|d)) ? ab*(c|d) ? A language is learnable in the limit if there exists a perfect child that generates only finitely many hypotheses.
SLIDE 51
Learning is not easy …
- Even simple languages like
regular languages are not learnable in the limit.
- Many settings: evil or well-
behaving mothers, with or without negative examples, frequencies, etc.
PAGE 50
sentence trace in event log language process model
SLIDE 52
PAGE 51
at the start of the century, process mining emerged as a new research topic remarkable progress over a relatively short period See keynote at Process Mining Camp 2013, http://fluxicon.com/camp/2013/
SLIDE 53
Process discovery challenge
(oversimplied no resources, data, etc.)
PAGE 52
a
start a = register request b = examine file c = check ticket d = decide e = reinitiate request f = send acceptance letter g = pay compensation h = send rejection letter
b c d g h e
end c1 c2 c3 c4 c5 t1
f
t2 t3 t4 t5 t6 t7 t8 t9 t10 t11 c6 c7 c8 c9 a,c,d,f,g a,b,c,d,e,c,d,g,f a,c,d,h
process model event log
SLIDE 54
Process discovery algorithms
(small selection)
PAGE 53
α algorithm α++ algorithm α# algorithm language-based regions state-based regions genetic mining heuristic mining hidden Markov models neural networks automata-based learning stochastic task graphs conformal process graph mining block structures multi-phase mining partial-order based mining fuzzy mining LTL mining ILP mining distributed genetic mining
SLIDE 55
Problem
PAGE 54
real process event data process model
record process discovery conformance checking process discovery conformance checking
“unknown” “only examples”
real process is unknown event logs covers only a fraction
- f all possible behavior
model needs to provide an abstraction: Murphy's Law of Process Mining
- nly positive examples
SLIDE 56
We only have example behavior (event log) and do not know the real process …
PAGE 55
M0 M0 M1
ideal or desired model based on perfect knowledge of real process
FN TP FP TN
descriptive or normative model (man-made or discovered)
L
event log regular behavior in log covered by model exceptional behavior in log covered by model exceptional behavior in log not covered by the model regular behavior in log not covered by model Problem II: in practice it is unclear where this line is Problem I: event log does not provide information about the whole universe of traces only a selected part
W.M.P. van der Aalst. Mediating Between Modeled and Observed Behavior: The Quest for the "Right" Process. In IEEE International Conference on Research Challenges in Information Science (RCIS 2013), pages 31-43. IEEE Computing Society, 2013.
SLIDE 57
Balance four forces
PAGE 56
fitness simplicity generalization precision
P r
- c
e s s M i n i n g
ability to explain
- bserved behavior
avoiding underfitting Occam’s Razor avoiding
- verfitting
lift gravity thrust drag
SLIDE 58
Example: one log four models
PAGE 57
acdeh abdeg adceh abdeh acdeg adceg adbeh acdefdbeh adbeg acdefbdeh acdefbdeg acdefdbeg adcefcdeh adcefdbeh adcefbdeg acdefbdefdbeg adcefdbeg adcefbdefbdeg adcefdbefbdeh adbefbdefdbeg adcefdbefcdefdbeg 455 191 177 144 111 82 56 47 38 33 14 11 9 8 5 3 2 2 1 1 1 # trace 1391
fitness simplicity generalization precision
Process Mining ability to explain- bserved behavior
- verfitting
… (all 21 variants seen in the log)
SLIDE 59
Model N1
PAGE 58
acdeh abdeg adceh abdeh acdeg adceg adbeh acdefdbeh adbeg acdefbdeh acdefbdeg acdefdbeg adcefcdeh adcefdbeh adcefbdeg acdefbdefdbeg adcefdbeg adcefbdefbdeg adcefdbefbdeh adbefbdefdbeg adcefdbefcdefdbeg 455 191 177 144 111 82 56 47 38 33 14 11 9 8 5 3 2 2 1 1 1 # trace 1391
a
start register request
b
examine thoroughly
c
examine casually
d
check ticket decide pay compensation reject request reinitiate request
e g h f
end
N1 : fitness = +, precision = +, generalization = +, simplicity = +
SLIDE 60
Model N2
PAGE 59
acdeh abdeg adceh abdeh acdeg adceg adbeh acdefdbeh adbeg acdefbdeh acdefbdeg acdefdbeg adcefcdeh adcefdbeh adcefbdeg acdefbdefdbeg adcefdbeg adcefbdefbdeg adcefdbefbdeh adbefbdefdbeg adcefdbefcdefdbeg 455 191 177 144 111 82 56 47 38 33 14 11 9 8 5 3 2 2 1 1 1 # trace 1391
a
start register request
c
examine casually
d
check ticket decide reject request
e h
end
N2 : fitness = -, precision = +, generalization = -, simplicity = +
SLIDE 61
Model N3
PAGE 60
acdeh abdeg adceh abdeh acdeg adceg adbeh acdefdbeh adbeg acdefbdeh acdefbdeg acdefdbeg adcefcdeh adcefdbeh adcefbdeg acdefbdefdbeg adcefdbeg adcefbdefbdeg adcefdbefbdeh adbefbdefdbeg adcefdbefcdefdbeg 455 191 177 144 111 82 56 47 38 33 14 11 9 8 5 3 2 2 1 1 1 # trace 1391
a
start register request
b
examine thoroughly
c
examine casually
d
check ticket decide pay compensation reject request reinitiate request
e g h f
end
N3 : fitness = +, precision = -, generalization = +, simplicity = +
SLIDE 62
Model N4
PAGE 61
acdeh abdeg adceh abdeh acdeg adceg adbeh acdefdbeh adbeg acdefbdeh acdefbdeg acdefdbeg adcefcdeh adcefdbeh adcefbdeg acdefbdefdbeg adcefdbeg adcefbdefbdeg adcefdbefbdeh adbefbdefdbeg adcefdbefcdefdbeg 455 191 177 144 111 82 56 47 38 33 14 11 9 8 5 3 2 2 1 1 1 # trace 1391 a
start register request
c
examine casually
d
check ticket decide reject request
e h
end
N4 : fitness = +, precision = +, generalization = -, simplicity = - a
register request
d
examine casually
c
check ticket decide reject request
e h a c
examine casually
d
check ticket decide
e g a d
examine casually
c
check ticket decide
e g
register request register request pay compensation pay compensation
a
register request
b d
check ticket decide reject request
e h a
register request
d b
check ticket decide reject request
e h a b d
check ticket decide
e g
register request pay compensation examine thoroughly examine thoroughly examine thoroughly
… (all 21 variants seen in the log)
SLIDE 63
PAGE 62
formal (not just a picture) fast (should not take years) sound (result should at least be free
- f deadlocks,
etc.) ability to balance all conformance dimensions (fitness, precision, generalization, and simplicity) incl. noise provide guarantees (not just a best effort) 1 2 3 4 5
SLIDE 64
PAGE 63
The future is bright, but how to get started? What are the main pitfalls of process modeling? What is process mining? Why is process discovery difficult? What are the main research challenges?
SLIDE 65
PAGE 64
Finding sheep with five legs
we are getting close…
SLIDE 66
PAGE 65
Distributing process mining problems to cope with big data
SLIDE 67
PAGE 66
On-the-fly process mining Operational support
SLIDE 68
Concept drift
PAGE 67
Concept drift
SLIDE 69
Cross-organizational mining
PAGE 68
cross-organizational / comparative process mining
SLIDE 70
PAGE 69
context aware process mining
SLIDE 71
PAGE 70
Suppor Supporting the pr ting the process
- cess
- f
- f pr
proces
- cess
s mining mining
SLIDE 72
PAGE 71
The future is bright, but how to get started? What are the main pitfalls of process modeling? What is process mining? Why is process discovery difficult? What are the main research challenges?
SLIDE 73
PAGE 72
SLIDE 74
The Sexiest Job of the 21st century (thanks to Moore's Law)
73
1965
SLIDE 75
PAGE 74
How to get started?
SLIDE 76
600+ plug-ins available covering the whole process mining spectrum
75
Download from: www.processmining.org
- pen-source (L-GPL)
SLIDE 77
Commercial Alternatives
- Disco (Fluxicon)
- Perceptive Process Mining
(before Futura Reflect and BPM|one)
- ARIS Process Performance
Manager
- QPR ProcessAnalyzer
- Interstage Process Discovery
(Fujitsu)
- Discovery Analyst (StereoLOGIC)
- XMAnalyzer (XMPro)
- …
76
SLIDE 78
How to Get Started?
Collect event data
- Minimal requirement:
events referring to an activity name and a process instance.
- Good to have:
timestamps, resource information, additional data elements.
- Challenges: scoping and
sometimes correlation. Collect questions
- What kind problems would
you like to address (cost, time, risk, compliance, service, etc.)?
- Related to discovery,
conformance, enhancement?
- Iterative process: can be
“curiosity driven” initially.
77
SLIDE 79
Conclusion
PAGE 78
process mining data-oriented analysis (data mining, machine learning, business intelligence) process model analysis (simulation, verification, etc.) performance-oriented questions, problems and solutions compliance-oriented questions, problems and solutions