[PPT] - Marlon Dumas University of Tartu, Estonia Estonian Theory Days

SLIDE 1

Marlon Dumas University of Tartu, Estonia

Estonian Theory Days | 3-4 October 2015

SLIDE 2

Process Mining

ü/û

event ¡log discovered ¡model Discovery Conformance Deviance Difference diagnos7cs Performance input ¡model Enhanced ¡model event ¡log’

2

SLIDE 3

Automated Process Discovery

3

Enter Loan Application Retrieve Applicant Data Compute Installments Approve Simple Application Approve Complex Application Notify Rejection Notify Eligibility

CID ¡ Task ¡ Time ¡Stamp ¡ … ¡ 13219 ¡Enter ¡Loan ¡Applica3on ¡ 2007-‑11-‑09 ¡T ¡11:20:10 ¡

‑ ¡

13219 ¡Retrieve ¡Applicant ¡Data ¡ 2007-‑11-‑09 ¡T ¡11:22:15 ¡

‑ ¡

13220 ¡Enter ¡Loan ¡Applica3on ¡ 2007-‑11-‑09 ¡T ¡11:22:40 ¡

‑ ¡

13219 ¡Compute ¡Installments ¡ 2007-‑11-‑09 ¡T ¡11:22:45 ¡

‑ ¡

13219 ¡No3fy ¡Eligibility ¡ 2007-‑11-‑09 ¡T ¡11:23:00 ¡

‑ ¡

13219 ¡Approve ¡Simple ¡Applica3on ¡ 2007-‑11-‑09 ¡T ¡11:24:30 ¡

‑ ¡

13220 ¡Compute ¡Installements ¡ 2007-‑11-‑09 ¡T ¡11:24:35 ¡

‑ ¡

… ¡ … ¡ … ¡ … ¡

SLIDE 4

Automated Process Discovery

Relations-based

– Alpha

4

SLIDE 5

Alpha Algorithm

Direct successors:

A > B, B > C, C > D, A > C, C > B, B > E, E > F C > E, E > G B > D

A B C D A C B E F

Causality:

A → B, C → D, A → C, B → E, C → E, E → F, E → G , B → D

Concurrency:

B ║ C

Exclusiveness: all other pairs

A B C E G A C B D

5

SLIDE 6

Alpha Relations Matrix

A B C D E F G A

# → → → # # #

B ←

# || → → # #

C ←

|| # → → # #

D

# ← ← # # # #

E

# ← ← # # → →

F

# # # # ← # #

G

# # # # ← # #

6

SLIDE 7

A B C D E F G A

# → → # # # #

B ←

# || → → # #

C ←

|| # → → # #

D

# ← ← # # # #

E

# ← ← # # → →

F

# # # # ← # #

G

# # # # ← # #

Alpha Algorithm – Patterns

7

⇔

a→ b, a→ c, b ║ c

SLIDE 8

Automated Process Discovery

Relations-based

– Alpha: lossy (Badouel 2012) – Alpha++, Alpha#, Alpha$ – Heuristics miner (frequency information)

Genetic
Region theory
Petri net synthesis
Integer Linear Programming (ILP)
…

8

SLIDE 9

Automated Process Discovery

Automated process discovery method Simplicity

minimal size & structural complexity

Precision

does not parse traces not in the log

Fitness

parses the traces of the log

Generalization

parses traces of the process not included in the log

9

SLIDE 10

Conformance Checking

?

10

SLIDE 11

Alignment-Based Conformance Check

Log Model

A B C D E A B B C

Alignment

E

Fitness Precision How much behavior of the log is captured by the model? How accurate is the model describing the log?

Munoz-Gama et al. Petri nets 2013 11

SLIDE 12

Imprecision of Alignment-Based Conformance Checking

{ABCD, ACBD} à 100%
{ABD, ACBCD} à 100%
{ACD, ABCBD} à 100%

12

SLIDE 13

Deviance Mining

13 T1 ¡<e11[d111:v111, ¡…, ¡d11n:v11n] ¡e12[d121:v121, ¡…, ¡d12m:v12m] ¡… ¡e1p[d1p1:v1p1, ¡…, ¡d1pm:v1pm]> ¡ … ¡ Tq ¡<eq1[dq11:vq11, ¡…, ¡dq1n:vq1n] ¡eq2[dq21:vq21, ¡…, ¡dq2m:vq2m] ¡… ¡eqp[dqp1:vqp1, ¡…, ¡dqpm:vqpm]> ¡ ¡

Find a function F: Trace à Boolean (or probability [0…1]) s.t.

F is an accurate approximation of the given labeling
F is explainable, e.g. set of simple predicates

SLIDE 14

Deviance Mining via Sequence Classification

Apply discriminative sequence mining methods to

extract features characteristic of one class

Build classification models (e.g. decision trees)
Extract difference diagnostics from classification model
C. Sun et al. Mining explicit rules for software process evaluation.

ICSSP’2013. 14

SLIDE 15

No Unified Foundation

Automated process discovery

Behavioral relations, theory of regions, ILP, …

Conformance checking

Replay, alignments

Deviance mining

Model delta analysis, sequence classification

≠

15

SLIDE 16

(Prime) Event Structures

Model of concurrency based on events

(occurrences of actions) and three relations

– Causality – Conflict – Concurrency

16

A B D E C

A C B D E E

SLIDE 17

Petri Nets à à Event Structures

17

τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ τ

a b c d d c b d d

SLIDE 18

Nets With Cycles à à Prefix Unfolding

18

Petri net N Complete prefix unfolding Causality-preservng prefix unfolding

SLIDE 19

Comparison of Event Structures

19

{},{} {A},{A} {A,B},{A,B} {A,B,D},{A,B,D} {A,B,C,D},{A,B,D} {A,B,C,D,E}, {A,B,D,E}

match(A ) match(B ) match(D ) left_hide( C) match(D )

A B D E C

A C B D E E

? ES1 ES2

Armas-Cervantes et al. Behavioral Comparison of Process Models Based on […] Event Structures. BPM’2014

Partially Synchronized Product (PSP) In ES1, tasks C and B are mutually exclusive, while in ES2, B precedes C

SLIDE 20

Event Structures for Process Mining

Deviance mining Conformance checking Process discovery

20

SLIDE 21

Event Logs è è Event Structures

B || C

Concurrency Oracle Run Merger

5 2 3

21

SLIDE 22

Event Structures for Log Delta Analysis

22 van Beest et al. Log delta analysis: Interpretable differencing of business process event logs. BPM’2015

SLIDE 23

Event Structures for Log Delta Analysis

In L1, task C can be skipped after B, whereas in L2 it cannot

van Beest et al. Log delta analysis: Interpretable differencing of business process event logs. BPM’2015 23

SLIDE 24

Log Delta Analysis

vs. Sequence Classification

448 cases 7329 events 363 cases, 7496 events Sequence classification 106-130 statements

Log delta analysis 48 statements

In L1, “Nursing Primary Assessment” is repeated after “Medical Assign Start” and “Triage Request”, while in L2 it is not. …

24 van Beest et al. Log delta analysis: Interpretable differencing of business process event logs. BPM’2015

SLIDE 25

Event Structures for Conformance Checking

Receive application Check credit history Assess loan risk Appraise property Assess eligibility

A B C D E

A B D E C

Receive application Appraise property Assess eligibility Check credit history Assess loan risk

A B C D E

A C B D E E

25

ABDE ADBE ACDE ADCE

SLIDE 26

Event Structures for Process Discovery?

26

ABDE ACDE ACDF Fold Merge Synth .

SLIDE 27

Process Mining Reloaded

27