Conformance
Checking Between
Designed and
Observed Processes
(My Viennese Christmas Sketches)
Artem Polyvyanyy
polyvyanyy.com | @uartem
1
Process Mining
Artem Polyvyanyy | Vienna, Austria | December 2018 2
Wil M. P. van der Aalst et al. Process Mining Manifesto. Business Process Management Workshops (1) 2011: 169-194
Quality Measures in Process Mining
Artem Polyvyanyy | Vienna, Austria | December 2018 3
3
- fitness
+ precision
- generalization
+ simplicity
+ fitness
+ precision
- generalization
+/- simplicity
+ fitness
- precision
+/- generalization
+/- simplicity
+ fitness
+ precision
+ generalization
+ simplicity
Adapted from Matthias Weidlich
Quality Measures in Process Mining
Artem Polyvyanyy | Vienna, Austria | December 2018 4
4
- fitness
+ precision
- generalization
+ simplicity
+ fitness
+ precision
- generalization
+/- simplicity
+ fitness
- precision
+/- generalization
+/- simplicity
+ fitness
+ precision
+ generalization
+ simplicity
Adapted from Matthias Weidlich
5
Artem Polyvyanyy
polyvyanyy.com | @uartem
Precision and Recall
for Process Mining
(under review)
Artem Polyvyanyy,
Andreas Solti,
Matthias Weidlich,
Claudio Di Ciccio, and
Jan Mendling
Precision and Recall
Artem Polyvyanyy | Vienna, Austria | December 2018 6
Precision: Fraction of the shared
behaviour relative to all
the behaviour of the model
Recall (a.k.a Fitness):
Fraction of the shared
behaviour relative to all
the behaviour of the log
Log Model
Log ∩ Model
m(Log ∩ Model)
m(Model)
m(Log ∩ Model)
m(Log)
How to define measure m and
operation ∩ ?
Which properties should
precision and recall have?
Imprecise Precisions
Artem Polyvyanyy | Vienna, Austria | December 2018 7
Niek Tax, Xixi Lu, Natalia Sidorova, Dirk Fahland, Wil M. P. van der Aalst:
The imprecisions of precision measures in process mining. Inf. Process. Lett. 135: 1-8 (2018)
A1: Precision is deterministic
A3: Precision between a log and a
flower model is the lowest
A4: Precision of a log on two language
equivalent models should be equal
A2:
A5:
m(L1 ∩ L3)
m(L3)
m(L2 ∩ L3)
m(L3)
≤
m(L1 ∩ L3)
m(L3)
m(L1 ∩ L2)
m(L2)
≤
Entropy-Based Precision and Recall
Artem Polyvyanyy | Vienna, Austria | December 2018 8
 The work co-authored by A. Polyvyanyy, A. Solti, M. Weidlich, C. Di Ciccio, and J. Mendling
 The only (to the best of my knowledge) precision measure that satisfies all the
proposed by the process mining community properties
 Under review, pre-pint available at: https://arxiv.org/pdf/1812.07334.pdf
 Based on the notion of topological entropy of an irreducible regular language, a language of
an ergodic, i.e., strongly connected, deterministic finite automaton (DFA)
 Let Cn(B) be the set of all words of length n in the language L(B) of a DFA B:
 Topological entropy can be computed as a Perron-Frobenius eigenvalue, i.e., a largest
eigenvalue, of the adjacency matrix of B; also if B describes an infinite language
 Logs and models get encoded as DFAs
 Model and log DFAs are short-circuited to obtain ergodic automata
 Overlap of DFAs is a well-defined operation and is used to compute the language in the
intersection of the model and log
 Minimization of DFAs is not required but empirically shown to speed up the computation of
eigenvalues
Precise Entropy-Based Precision
Artem Polyvyanyy | Vienna, Austria | December 2018 9
A1: Precision is deterministic
A3: Precision between a log and a
flower model is the lowest
A4: Precision of a log on two language
equivalent models should be equal
A2:
A5:
ent (L1 ∩ L3)
ent(L3)
ent(L2 ∩ L3)
ent(L3)
<
ent (L1 ∩ L3)
ent(L3)
ent(L1 ∩ L2)
ent(L2)
<
Entropy-based precision
note the strict monotonicity;
also fulfils additional properties
(to be shown in a few slides)
Precise Entropy-Based Precision
Artem Polyvyanyy | Vienna, Austria | December 2018 10
a{<min>,<max>} is short-hand for enumerating the
minimal and maximal number of repetitions of a, e.g.,
a{0,2}○b encodes language [<b>,<a,b>,<a,a,b>].
Precise Entropy-Based Precision
Artem Polyvyanyy | Vienna, Austria | December 2018 11
Precise Entropy-Based Precision
Artem Polyvyanyy | Vienna, Austria | December 2018 12
fits model
Precise Entropy-Based Precision
Artem Polyvyanyy | Vienna, Austria | December 2018 13
fits model
Precise Entropy-Based Precision
Artem Polyvyanyy | Vienna, Austria | December 2018 14
ATAED’18
recent precent
✓ ✓
✓ ✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Conjecture
Undefined for
infinite languages
Difficult
case
A1
A4
~A5
Yet Another Precision Measure
Artem Polyvyanyy | Vienna, Austria | December 2018 15
 Uses the k-th order Markovian abstraction
 Claims fulfilment of the 5 “weak” axioms by Tax et al.
(i.e., A2 and A5 are based on ≤)
 Parameter k for which axioms hold is specific for a
given set of models and logs
 Gets computationally demanding as k gets large
 A k-th order Markovian abstraction can be seen as a
behavioural profile
 Behavioral profiles are less expressive than regular
languages, refer to A.Polyvyanyy et al.: On the
expressive power of behavioral profiles. Formal Asp.
Comput. 28(4): 597-613 (2016)
 Measures for comparing behaviours based on beha-
vioural profiles were first proposed in M.Weidlich et al.:
Process compliance analysis based on behavioural
profiles. Inf. Syst. 36(7):1009-1025 (2011)
M1-abstraction M2-abstraction
BPM’18
Entropy-Based Recall
Artem Polyvyanyy | Vienna, Austria | December 2018 16
 Given a sequential model of 10 activities and a fitting event log with no noise, we
gradually increase the amount of noisy traces in the event log
 Noise is defined as removing, adding, or swapping events in the event log
Implementation and Evaluation
Artem Polyvyanyy | Vienna, Austria | December 2018 17
 Event logs are available via: https://data.4tu.nl/repository/collection:event_logs_real
 Eigenvalue computation implementation available via:
https://github.com/andreas-solti/matrix-toolkits-java
 Experiments and implementation available via:
https://github.com/andreas-solti/eigen-measure
Source code:
Scalability evaluation setup:
 For each log, mined a model with InductiveMiner (noise threshold 0.2)
 In all but two of the real-life cases our method computes the automata and their largest
eigenvalues within 10 minutes to yield the precision and recall values
Conclusion and Future Work
Artem Polyvyanyy | Vienna, Austria | December 2018 18
 The only (to the best of my knowledge) precision measure that satisfies all
the proposed by the process mining community properties
 Under review, pre-pint available at: https://arxiv.org/pdf/1812.07334.pdf
Possible avenues for future work include:
 Extend the measures to support partial trace matching (in progress)
 Extend the measures to support infinite-state behaviors (initial idea available)
 Improve efficiency (use efficient implementations/optimizations)
 Extend the measures to support frequencies of traces (under review)
 Applications of the measure in Process Mining and other areas,
e.g. Software Engineering
Limitations:
 Computation efficiency can be improved (though already feasible)
 Partial trace matching is not supported (desired properties of such measures
are yet to be understood)
 Infinite-state behaviors are not supported
 Frequencies of traces are not accounted for (desired properties of such
measures are yet to be understood)
Other Interests
Artem Polyvyanyy | Vienna, Austria | December 2018 19
 Process Querying including languages for process querying.
 Series of workshops: http://processquerying.com/
 Process Query Language (PQL): https://github.com/processquerying/PQL
 Process Querying Framework: Artem Polyvyanyy, Chun Ouyang, Alistair Barros,
Wil M. P. van der Aalst: Process querying: Enabling business intelligence through
query-based process analytics. Decision Support Systems 100: 41-56 (2017)
 Edited book project
 Process Forecasting: Rouven Poll, Artem Polyvyanyy, Michael Rosemann, Maximilian
Röglinger, Lea Rupprecht: Process Forecasting: Towards Proactive Business Process
Management. BPM 2018: 496-512
 Stochastic Process Mining: Work on stochastic precision and recall with Sander Leemans
 Information Systems Modeling: Work on a formalism for encoding processes and data
with Jan Martijn van der Werf:
 Language for modeling information systems for which reachability is decidable
http://www.cs.uu.nl/research/techreps/repo/CS-2018/2018-004.pdf
 Isotactics: Artem Polyvyanyy, Jan Sürmeli, Matthias Weidlich: Interleaving isotactics - An
equivalence notion on behaviour abstractions. Theor. Comput. Sci. 737: 1-18 (2018)
Process Querying Framework
Artem Polyvyanyy | Vienna, Austria | December 2018 20
Edited book entitled
“Process Querying Methods”
 To appear in mid-2019!
 53 contributors: Ahmed Awad, Alexander Artikis, Amal Elgammal, Amal Fawzy
Elgammal, Amin Beheshti, Andreas Oberweis, Andreas Schoknecht, Antonia M.
Reina, Antonio Cancela, Artem Polyvyanyy, Boualem Benatallah, Carl Corea, Chiara Di
Francescomarino, Christoph Drodt, David Knuplesch, Dennis Riehle, Eduardo
González López de Murillas, Emiliano Reynares, Fabrizio Smith, Farhad Amouzgar,
Francesco Taglino, Hajo A. Reijers, Hamid R. Motahari Nezhad, Han van der Aa,
Harald Störrle, Jianmin Wang, Jorge Roa, Jose Miguel Pérez, Kazimierz Subieta, Klaus
Kammerer, Luisa Parody, Manfred Reichert, María Laura Caliusco, María Teresa
Gómez, Mariusz Momotko, Matthias Weidlich, Maurizio Proietti, Oktay Turetken,
Pablo Villarreal, Paolo Tonella, Patrick Delfmann, Peter Fettke, Ralf Laue, Remco
Dijkman, Rik Eshuis, Rüdiger Pryss, Samira Ghodratnama, Stefanie Rinderle-Ma,
Steffen Höhenberger, Tao Jin, Tom Thaler, Vlad Acretoaie, and Wil M. P. van der Aalst.
21
Isotactics
Artem Polyvyanyy | Vienna, Austria | December 2018 22
23
Impact-Driven
Process Model
Repair
Artem Polyvyanyy,
Wil M. P. van der Aalst,
Arthur H. M. ter Hofstede, and
Moe T. Wynn
Artem Polyvyanyy
polyvyanyy.com | @uartem
Design vs Experience
Artem Polyvyanyy | Vienna, Austria | December 2018 24
Conformance Checking
Artem Polyvyanyy | Vienna, Austria | December 2018 25
a
p1
t1 p2
b c
d e
p3
p4
p5
t2 t3
t4 t5
a b c
a b c
t1 t2 t3
γ1 =
a d e
a d e
t1 t4 t5
γ2 =
perfect alignments
a b x
a b
t1 t2
γ3 =
real-world traces
designed traces
e ≫
c
t3
≫ ≫
a b x
a b
t1 t2
γ4 =
e≫
c
t3
≫ ≫
a b
a
t1 t4
γ5 = ≫d
≫ x e
e
t5
≫
0 0 1 1 2 0 0 1 2 1 0 3 1 1 0+ + + + + + + + == =4 4 5+ + + +
move on trace move on modelsynchronous move
cost(γ3)=cost(γ4)=4 and cost(γ5)=5; γ3 and γ4 are optimal alignments!
Process Model Repair
Artem Polyvyanyy | Vienna, Austria | December 2018 26
a
p1
t1 p2
b c
d e
p3
p4
p5
t2 t3
t4 t5
a b x
a b
t1 t2
γ4 =
e≫
c
t3
≫ ≫
0 0 1 2 1
γ6 =
a e
a e
t1 t5
≫
d
t4
0 3 0
 = 7R=({x},{d})
insert skip
x t6
t7
γ7 =
a
a
t1
0
b
b
t2
0
x
x
t6
0
c
t3
2
≫
1
e
≫ γ9 =
a
a
t1
0
≫
t7
0
b
≫
1
≫
1
x
0
e
e
t5
γ8 =
a
a
t1
0
≫
t7
0 0
e
e
t5
+ + + + = 3 + + + + = 2 + + = 0
 = 2
Note the problem with x!
ττ
γ8 and γ9 are optimal alignments!
Towards Optimal Repair
Artem Polyvyanyy | Vienna, Austria | December 2018 27
R=({x},{d})
a
p1
t1 p2
b c
d e
p3
p4
p5
t2 t3
t4 t5
γ10 =
a
a
t1
0
d
t4
0
≫
1
b
≫
0
x
≫
e
e
t5
0
 = 1
=
a e
a e
t1 t5
≫
d
t4
0 0 0
γ11
simulating repair via “free” moves
a
p1
t1 p2
b c
d e
p3
p4
p5
t2 t3
t4 t5
t7
x t6
 = 1
γ12 =
a
a
t1
0
≫
t7
0
b
≫
1 0
e
e
t5
+ + + + = 1
x
x
t6
0
γ8 =
a
a
t1
0
≫
t7
0 0
e
e
t5
+ + = 0
ττ
γ8 and γ12 are optimal alignments!
Note no problem with x!
Optimal Repair (Theorem 5.3)
Artem Polyvyanyy | Vienna, Austria | December 2018 28
There exist optimal alignments between
the traces and the repaired model which:
(i) Demonstrate that the traces fit the
repaired model “at least as good” as they
fit the original model;
(ii) “Fulfil” the repair recommendation,
e.g., γ8 and γ12 contain no “bad” moves
(x,≫) and (≫,d).
a
p1
t1 p2
b c
d e
p3
p4
p5
t2 t3
t4 t5
a
p1
t1 p2
b c
d e
p3
p4
p5
t2 t3
t4 t5
t7
x t6
 = 1
γ12 =
a
a
t1
0
≫
t7
0
b
≫
1 0
e
e
t5
+ + + + = 1
x
x
t6
0
γ8 =
a
a
t1
0
≫
t7
0 0
e
e
t5
+ + = 0
ττ
R=({x},{d})
γ8 and γ12 are optimal alignments!
Note no problem with x!
One More Example
Artem Polyvyanyy | Vienna, Austria | December 2018 29
a
p1
c
d
b f
g
h
e
t1
t2
t3
t4
t5
t6
t7
t8
t9
t10
t11
p2
p3
p4 p5
p6
p7
p8
p9
p10 p11
10 x
9 x
6 x
2 x
9 x
7 x
2 x
 = 120
All deviations
cost one Peso!
A Possible Repair
Artem Polyvyanyy | Vienna, Austria | December 2018 30
R = ({f,e,x},{c,f,g})
 = 47
insert skip
a
p1
c
d
b
f
g
he
t1
t2
t3
t4
t5
t6
t7
t8
t9
t10
t11
p2 p3
p4 p5 p6
p7
p8
p9
p10 p11
t12
t13
t14
x
f
e
t15
t16 t17
Another Possible Repair
Artem Polyvyanyy | Vienna, Austria | December 2018 31
R = ({f,e},{g,d,e,c})
 = 33
insert skip
a
p1
c
d
b
f
g
h
e
t1
t2
t3
t4
t5
t6
t7
t8
t9
t10
t11
p2
p3
p4
p5
p6
p7
p8
p9
p10
p11
t13
t15
t16
t17
t18
e f
t12 t13
et14
Towards Optimal Repair Recommendation
Artem Polyvyanyy | Vienna, Austria | December 2018 32
R = ({f,x},{d,e,c,h})
 = 25
insert skip
a
p1
c
d
b
f
g
he
t1
t2
t3
t4
t5
t6
t7
t8
t9
t10
t11
p2
p3
p4
p5
p6
p7
p8
p9
p10
p11
t13
t12
f
x
t14
f
t15
t16
t17
t18
costs of repair
recommendation
implemented
heuristics
numbers of discovered
recommendations
numbers of alignment
computations deviation costs
Conclusion and Future Work
Artem Polyvyanyy | Vienna, Austria | December 2018 33
In this work, we …
 Defined, and proposed a solution to, the optimal process model repair problem
 Defined, and proposed several (approximate) solutions to, the optimal repair
recommendation problem
 Developed a publicly available tool that implements all the invented techniques,
https://svn.win.tue.nl/repos/prom/Packages/SelectivePRepair/Trunk/release/
 Reported on lessons learned from experiments with real-world data, e.g.,
for small repairs proposed heuristics perform just as well as brute-force
Possible avenues for future work include:
 Case studies to better understand requirements of process model repair
 Aesthetic process model repair
 Repair heuristics with guarantees
 Repairs based on other conformance data structures
 Repairs that go beyond trace fitness
 Applications of the proposed repair technique

Conformance checking between designed and observed processes

  • 1.
    Conformance Checking Between Designed and ObservedProcesses (My Viennese Christmas Sketches) Artem Polyvyanyy polyvyanyy.com | @uartem 1
  • 2.
    Process Mining Artem Polyvyanyy| Vienna, Austria | December 2018 2 Wil M. P. van der Aalst et al. Process Mining Manifesto. Business Process Management Workshops (1) 2011: 169-194
  • 3.
    Quality Measures inProcess Mining Artem Polyvyanyy | Vienna, Austria | December 2018 3 3 - fitness + precision - generalization + simplicity + fitness + precision - generalization +/- simplicity + fitness - precision +/- generalization +/- simplicity + fitness + precision + generalization + simplicity Adapted from Matthias Weidlich
  • 4.
    Quality Measures inProcess Mining Artem Polyvyanyy | Vienna, Austria | December 2018 4 4 - fitness + precision - generalization + simplicity + fitness + precision - generalization +/- simplicity + fitness - precision +/- generalization +/- simplicity + fitness + precision + generalization + simplicity Adapted from Matthias Weidlich
  • 5.
    5 Artem Polyvyanyy polyvyanyy.com |@uartem Precision and Recall for Process Mining (under review) Artem Polyvyanyy, Andreas Solti, Matthias Weidlich, Claudio Di Ciccio, and Jan Mendling
  • 6.
    Precision and Recall ArtemPolyvyanyy | Vienna, Austria | December 2018 6 Precision: Fraction of the shared behaviour relative to all the behaviour of the model Recall (a.k.a Fitness): Fraction of the shared behaviour relative to all the behaviour of the log Log Model Log ∩ Model m(Log ∩ Model) m(Model) m(Log ∩ Model) m(Log) How to define measure m and operation ∩ ? Which properties should precision and recall have?
  • 7.
    Imprecise Precisions Artem Polyvyanyy| Vienna, Austria | December 2018 7 Niek Tax, Xixi Lu, Natalia Sidorova, Dirk Fahland, Wil M. P. van der Aalst: The imprecisions of precision measures in process mining. Inf. Process. Lett. 135: 1-8 (2018) A1: Precision is deterministic A3: Precision between a log and a flower model is the lowest A4: Precision of a log on two language equivalent models should be equal A2: A5: m(L1 ∩ L3) m(L3) m(L2 ∩ L3) m(L3) ≤ m(L1 ∩ L3) m(L3) m(L1 ∩ L2) m(L2) ≤
  • 8.
    Entropy-Based Precision andRecall Artem Polyvyanyy | Vienna, Austria | December 2018 8  The work co-authored by A. Polyvyanyy, A. Solti, M. Weidlich, C. Di Ciccio, and J. Mendling  The only (to the best of my knowledge) precision measure that satisfies all the proposed by the process mining community properties  Under review, pre-pint available at: https://arxiv.org/pdf/1812.07334.pdf  Based on the notion of topological entropy of an irreducible regular language, a language of an ergodic, i.e., strongly connected, deterministic finite automaton (DFA)  Let Cn(B) be the set of all words of length n in the language L(B) of a DFA B:  Topological entropy can be computed as a Perron-Frobenius eigenvalue, i.e., a largest eigenvalue, of the adjacency matrix of B; also if B describes an infinite language  Logs and models get encoded as DFAs  Model and log DFAs are short-circuited to obtain ergodic automata  Overlap of DFAs is a well-defined operation and is used to compute the language in the intersection of the model and log  Minimization of DFAs is not required but empirically shown to speed up the computation of eigenvalues
  • 9.
    Precise Entropy-Based Precision ArtemPolyvyanyy | Vienna, Austria | December 2018 9 A1: Precision is deterministic A3: Precision between a log and a flower model is the lowest A4: Precision of a log on two language equivalent models should be equal A2: A5: ent (L1 ∩ L3) ent(L3) ent(L2 ∩ L3) ent(L3) < ent (L1 ∩ L3) ent(L3) ent(L1 ∩ L2) ent(L2) < Entropy-based precision note the strict monotonicity; also fulfils additional properties (to be shown in a few slides)
  • 10.
    Precise Entropy-Based Precision ArtemPolyvyanyy | Vienna, Austria | December 2018 10 a{<min>,<max>} is short-hand for enumerating the minimal and maximal number of repetitions of a, e.g., a{0,2}○b encodes language [<b>,<a,b>,<a,a,b>].
  • 11.
    Precise Entropy-Based Precision ArtemPolyvyanyy | Vienna, Austria | December 2018 11
  • 12.
    Precise Entropy-Based Precision ArtemPolyvyanyy | Vienna, Austria | December 2018 12 fits model
  • 13.
    Precise Entropy-Based Precision ArtemPolyvyanyy | Vienna, Austria | December 2018 13 fits model
  • 14.
    Precise Entropy-Based Precision ArtemPolyvyanyy | Vienna, Austria | December 2018 14 ATAED’18 recent precent ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ Conjecture Undefined for infinite languages Difficult case A1 A4 ~A5
  • 15.
    Yet Another PrecisionMeasure Artem Polyvyanyy | Vienna, Austria | December 2018 15  Uses the k-th order Markovian abstraction  Claims fulfilment of the 5 “weak” axioms by Tax et al. (i.e., A2 and A5 are based on ≤)  Parameter k for which axioms hold is specific for a given set of models and logs  Gets computationally demanding as k gets large  A k-th order Markovian abstraction can be seen as a behavioural profile  Behavioral profiles are less expressive than regular languages, refer to A.Polyvyanyy et al.: On the expressive power of behavioral profiles. Formal Asp. Comput. 28(4): 597-613 (2016)  Measures for comparing behaviours based on beha- vioural profiles were first proposed in M.Weidlich et al.: Process compliance analysis based on behavioural profiles. Inf. Syst. 36(7):1009-1025 (2011) M1-abstraction M2-abstraction BPM’18
  • 16.
    Entropy-Based Recall Artem Polyvyanyy| Vienna, Austria | December 2018 16  Given a sequential model of 10 activities and a fitting event log with no noise, we gradually increase the amount of noisy traces in the event log  Noise is defined as removing, adding, or swapping events in the event log
  • 17.
    Implementation and Evaluation ArtemPolyvyanyy | Vienna, Austria | December 2018 17  Event logs are available via: https://data.4tu.nl/repository/collection:event_logs_real  Eigenvalue computation implementation available via: https://github.com/andreas-solti/matrix-toolkits-java  Experiments and implementation available via: https://github.com/andreas-solti/eigen-measure Source code: Scalability evaluation setup:  For each log, mined a model with InductiveMiner (noise threshold 0.2)  In all but two of the real-life cases our method computes the automata and their largest eigenvalues within 10 minutes to yield the precision and recall values
  • 18.
    Conclusion and FutureWork Artem Polyvyanyy | Vienna, Austria | December 2018 18  The only (to the best of my knowledge) precision measure that satisfies all the proposed by the process mining community properties  Under review, pre-pint available at: https://arxiv.org/pdf/1812.07334.pdf Possible avenues for future work include:  Extend the measures to support partial trace matching (in progress)  Extend the measures to support infinite-state behaviors (initial idea available)  Improve efficiency (use efficient implementations/optimizations)  Extend the measures to support frequencies of traces (under review)  Applications of the measure in Process Mining and other areas, e.g. Software Engineering Limitations:  Computation efficiency can be improved (though already feasible)  Partial trace matching is not supported (desired properties of such measures are yet to be understood)  Infinite-state behaviors are not supported  Frequencies of traces are not accounted for (desired properties of such measures are yet to be understood)
  • 19.
    Other Interests Artem Polyvyanyy| Vienna, Austria | December 2018 19  Process Querying including languages for process querying.  Series of workshops: http://processquerying.com/  Process Query Language (PQL): https://github.com/processquerying/PQL  Process Querying Framework: Artem Polyvyanyy, Chun Ouyang, Alistair Barros, Wil M. P. van der Aalst: Process querying: Enabling business intelligence through query-based process analytics. Decision Support Systems 100: 41-56 (2017)  Edited book project  Process Forecasting: Rouven Poll, Artem Polyvyanyy, Michael Rosemann, Maximilian Röglinger, Lea Rupprecht: Process Forecasting: Towards Proactive Business Process Management. BPM 2018: 496-512  Stochastic Process Mining: Work on stochastic precision and recall with Sander Leemans  Information Systems Modeling: Work on a formalism for encoding processes and data with Jan Martijn van der Werf:  Language for modeling information systems for which reachability is decidable http://www.cs.uu.nl/research/techreps/repo/CS-2018/2018-004.pdf  Isotactics: Artem Polyvyanyy, Jan Sürmeli, Matthias Weidlich: Interleaving isotactics - An equivalence notion on behaviour abstractions. Theor. Comput. Sci. 737: 1-18 (2018)
  • 20.
    Process Querying Framework ArtemPolyvyanyy | Vienna, Austria | December 2018 20
  • 21.
    Edited book entitled “ProcessQuerying Methods”  To appear in mid-2019!  53 contributors: Ahmed Awad, Alexander Artikis, Amal Elgammal, Amal Fawzy Elgammal, Amin Beheshti, Andreas Oberweis, Andreas Schoknecht, Antonia M. Reina, Antonio Cancela, Artem Polyvyanyy, Boualem Benatallah, Carl Corea, Chiara Di Francescomarino, Christoph Drodt, David Knuplesch, Dennis Riehle, Eduardo González López de Murillas, Emiliano Reynares, Fabrizio Smith, Farhad Amouzgar, Francesco Taglino, Hajo A. Reijers, Hamid R. Motahari Nezhad, Han van der Aa, Harald Störrle, Jianmin Wang, Jorge Roa, Jose Miguel Pérez, Kazimierz Subieta, Klaus Kammerer, Luisa Parody, Manfred Reichert, María Laura Caliusco, María Teresa Gómez, Mariusz Momotko, Matthias Weidlich, Maurizio Proietti, Oktay Turetken, Pablo Villarreal, Paolo Tonella, Patrick Delfmann, Peter Fettke, Ralf Laue, Remco Dijkman, Rik Eshuis, Rüdiger Pryss, Samira Ghodratnama, Stefanie Rinderle-Ma, Steffen Höhenberger, Tao Jin, Tom Thaler, Vlad Acretoaie, and Wil M. P. van der Aalst. 21
  • 22.
    Isotactics Artem Polyvyanyy |Vienna, Austria | December 2018 22
  • 23.
    23 Impact-Driven Process Model Repair Artem Polyvyanyy, WilM. P. van der Aalst, Arthur H. M. ter Hofstede, and Moe T. Wynn Artem Polyvyanyy polyvyanyy.com | @uartem
  • 24.
    Design vs Experience ArtemPolyvyanyy | Vienna, Austria | December 2018 24
  • 25.
    Conformance Checking Artem Polyvyanyy| Vienna, Austria | December 2018 25 a p1 t1 p2 b c d e p3 p4 p5 t2 t3 t4 t5 a b c a b c t1 t2 t3 γ1 = a d e a d e t1 t4 t5 γ2 = perfect alignments a b x a b t1 t2 γ3 = real-world traces designed traces e ≫ c t3 ≫ ≫ a b x a b t1 t2 γ4 = e≫ c t3 ≫ ≫ a b a t1 t4 γ5 = ≫d ≫ x e e t5 ≫ 0 0 1 1 2 0 0 1 2 1 0 3 1 1 0+ + + + + + + + == =4 4 5+ + + + move on trace move on modelsynchronous move cost(γ3)=cost(γ4)=4 and cost(γ5)=5; γ3 and γ4 are optimal alignments!
  • 26.
    Process Model Repair ArtemPolyvyanyy | Vienna, Austria | December 2018 26 a p1 t1 p2 b c d e p3 p4 p5 t2 t3 t4 t5 a b x a b t1 t2 γ4 = e≫ c t3 ≫ ≫ 0 0 1 2 1 γ6 = a e a e t1 t5 ≫ d t4 0 3 0  = 7R=({x},{d}) insert skip x t6 t7 γ7 = a a t1 0 b b t2 0 x x t6 0 c t3 2 ≫ 1 e ≫ γ9 = a a t1 0 ≫ t7 0 b ≫ 1 ≫ 1 x 0 e e t5 γ8 = a a t1 0 ≫ t7 0 0 e e t5 + + + + = 3 + + + + = 2 + + = 0  = 2 Note the problem with x! ττ γ8 and γ9 are optimal alignments!
  • 27.
    Towards Optimal Repair ArtemPolyvyanyy | Vienna, Austria | December 2018 27 R=({x},{d}) a p1 t1 p2 b c d e p3 p4 p5 t2 t3 t4 t5 γ10 = a a t1 0 d t4 0 ≫ 1 b ≫ 0 x ≫ e e t5 0  = 1 = a e a e t1 t5 ≫ d t4 0 0 0 γ11 simulating repair via “free” moves a p1 t1 p2 b c d e p3 p4 p5 t2 t3 t4 t5 t7 x t6  = 1 γ12 = a a t1 0 ≫ t7 0 b ≫ 1 0 e e t5 + + + + = 1 x x t6 0 γ8 = a a t1 0 ≫ t7 0 0 e e t5 + + = 0 ττ γ8 and γ12 are optimal alignments! Note no problem with x!
  • 28.
    Optimal Repair (Theorem5.3) Artem Polyvyanyy | Vienna, Austria | December 2018 28 There exist optimal alignments between the traces and the repaired model which: (i) Demonstrate that the traces fit the repaired model “at least as good” as they fit the original model; (ii) “Fulfil” the repair recommendation, e.g., γ8 and γ12 contain no “bad” moves (x,≫) and (≫,d). a p1 t1 p2 b c d e p3 p4 p5 t2 t3 t4 t5 a p1 t1 p2 b c d e p3 p4 p5 t2 t3 t4 t5 t7 x t6  = 1 γ12 = a a t1 0 ≫ t7 0 b ≫ 1 0 e e t5 + + + + = 1 x x t6 0 γ8 = a a t1 0 ≫ t7 0 0 e e t5 + + = 0 ττ R=({x},{d}) γ8 and γ12 are optimal alignments! Note no problem with x!
  • 29.
    One More Example ArtemPolyvyanyy | Vienna, Austria | December 2018 29 a p1 c d b f g h e t1 t2 t3 t4 t5 t6 t7 t8 t9 t10 t11 p2 p3 p4 p5 p6 p7 p8 p9 p10 p11 10 x 9 x 6 x 2 x 9 x 7 x 2 x  = 120 All deviations cost one Peso!
  • 30.
    A Possible Repair ArtemPolyvyanyy | Vienna, Austria | December 2018 30 R = ({f,e,x},{c,f,g})  = 47 insert skip a p1 c d b f g he t1 t2 t3 t4 t5 t6 t7 t8 t9 t10 t11 p2 p3 p4 p5 p6 p7 p8 p9 p10 p11 t12 t13 t14 x f e t15 t16 t17
  • 31.
    Another Possible Repair ArtemPolyvyanyy | Vienna, Austria | December 2018 31 R = ({f,e},{g,d,e,c})  = 33 insert skip a p1 c d b f g h e t1 t2 t3 t4 t5 t6 t7 t8 t9 t10 t11 p2 p3 p4 p5 p6 p7 p8 p9 p10 p11 t13 t15 t16 t17 t18 e f t12 t13 et14
  • 32.
    Towards Optimal RepairRecommendation Artem Polyvyanyy | Vienna, Austria | December 2018 32 R = ({f,x},{d,e,c,h})  = 25 insert skip a p1 c d b f g he t1 t2 t3 t4 t5 t6 t7 t8 t9 t10 t11 p2 p3 p4 p5 p6 p7 p8 p9 p10 p11 t13 t12 f x t14 f t15 t16 t17 t18 costs of repair recommendation implemented heuristics numbers of discovered recommendations numbers of alignment computations deviation costs
  • 33.
    Conclusion and FutureWork Artem Polyvyanyy | Vienna, Austria | December 2018 33 In this work, we …  Defined, and proposed a solution to, the optimal process model repair problem  Defined, and proposed several (approximate) solutions to, the optimal repair recommendation problem  Developed a publicly available tool that implements all the invented techniques, https://svn.win.tue.nl/repos/prom/Packages/SelectivePRepair/Trunk/release/  Reported on lessons learned from experiments with real-world data, e.g., for small repairs proposed heuristics perform just as well as brute-force Possible avenues for future work include:  Case studies to better understand requirements of process model repair  Aesthetic process model repair  Repair heuristics with guarantees  Repairs based on other conformance data structures  Repairs that go beyond trace fitness  Applications of the proposed repair technique