Chapter 8Mining Additional Perspectives

prof.dr.ir. Wil van der Aalstwww.processmining.org

http://www.layoutsparks.com/myspace-layouts/footsteps+in++s_0�

Overview

PAGE 1

Part I: Preliminaries

Chapter 2 Process Modeling and Analysis

Chapter 3Data Mining

Part II: From Event Logs to Process Models

Chapter 4 Getting the Data

Chapter 5 Process Discovery: An Introduction

Chapter 6 Advanced Process Discovery Techniques

Part III: Beyond Process Discovery

Chapter 7 Conformance Checking

Chapter 8 Mining Additional Perspectives

Chapter 9 Operational Support

Part IV: Putting Process Mining to Work

Chapter 10 Tool Support

Chapter 11 Analyzing “Lasagna Processes”

Chapter 12 Analyzing “Spaghetti Processes”

Part V: Reflection

Chapter 13Cartography and Navigation

Chapter 14Epilogue

Chapter 1 Introduction

Mining additional perspectives(one type of enhancement, cf. repair in context of conformance checking)

PAGE 2

software system

(process)model

eventlogs

modelsanalyzes

discovery

records events, e.g., messages,

transactions, etc.

specifies configures implements

analyzes

supports/controls

enhancement

conformance

“world”

people machines

organizationscomponents

businessprocesses

Replay: Connecting events to model elements is essential for process mining

PAGE 3

event log process model

Play-In

event logprocess model

Play-Out

event log process model

Replay

• extended model showing times, frequencies, etc.

• diagnostics• predictions• recommendations

A

B

C

DE

p2

end

p4

p3p1

start

Remember: Replay!

PAGE 4

A B C D

A

B

C

DE

p2

end

p4

p3p1

start

Replay can detect problems

PAGE 5

AC D

Problem!missing token

Problem!token left behind

A

B

C

DE

p2

end

p4

p3p1

start

Replay can extract timing information

PAGE 6

A5B8 C9 D13

5

8

9

13

3

4

5

43

265

8

764

7

74

3

A

B

C

DE

p2

end

p4

p3p1

start

Decision mining: “Red” cases

PAGE 7

A B C D

A

B

C

DE

p2

end

p4

p3p1

start

Decision mining: “Blue” cases

PAGE 8

A E D If red then B+C;If blue then E;

Starting point: connected event log and model

PAGE 9

...

process case

activity activity instance

event

attribute

timestamp

resource*

1

*1

*

1

*1

1

**1

1

*a

b

c

d

e

f g

h

i

event level

costs

trans- action

model level

instance level

j

k

Process

PAGE 10

software system

(process)model

eventlogs

modelsanalyzes

discovery

records events, e.g., messages,

transactions, etc.

specifies configures implements

analyzes

supports/controls

enhancement

conformance

“world”

people machines

organizationscomponents

businessprocessesthe initial

process model is made by

hand or discovered from

the event logevents have

attributes relating to

various perspectives

conformance checking is used to relate the initial

model and event log

1

2

3

4

the model is extended using the additional information in the

event log

5

integrated model showing multiple

perspectives

Attributes in event logs

PAGE 11

Cases may also have attributes

PAGE 12

Helicopter view: Dotted charts

PAGE 13

time can be absolute or relative and real or logical

time

class

each line corresponds to a class, e.g., a case, a

resource, a customer, or an activity

each dot corresponds to an event

the color and shape of a dot may depend on attributes of the

event

activity : decidetype : starttime : 06-01-2011:11.18resource : Saracost : -custid : 9911name : Smithtype : goldregion : southamount : 989.50

Dotted chart for a process of a housing agency using absolute time

PAGE 14

Zooming in

PAGE 15

Same log, relative time

PAGE 16

Organizational mining

PAGE 17

Resource-activity matrix

PAGE 18

mean number of times a resource performs an activity per case

Activity a is executed exactly once for each case (take the sum of the first column). Pete, Mike, and Ellen are the only ones executing this activity. In 30% of the cases, a is executed by Pete, 50% is executed by Pete, and 20% is executed by Ellen. Activities e and f are always executed by Sara. Activity e is executed, on average, 2.3 times per case. Etc.

Social network analysis

PAGE 19

relationship

organizational entity (resource, person, role, department, etc.)

x z

y the size of the oval indicates the weight of the entity

the thickness of the arc indicates the weight of the relationship

w=0.90

w=0.30 w=0.35

w=0.98

w=0.80

w=0.15

w=0.08

Handover of work matrix

PAGE 20

Count the number of times work is handed over from one resource to another (on average per case).

The causal dependencies in the process model are used to count handovers in the event log.

Social network based on handover of work (threshold of 0.1)

PAGE 21

Pete

Mike

Ellen

Sue

Sean

Sara

In this figure only the thickness of the arcs is based on frequencies.

Handover of work at role level

PAGE 22

w=1.5

w=0.5

w=3.45

w=1.15

w=2,95

w=0.65w=1.3

Assistantw=5.45

Expertw=1.15

Managerw=3.6In this figure also

the size of each node is based on frequencies.

Profile

PAGE 23

Social network based on similarity of profiles

PAGE 24

Pete

Mike Ellen

SueSean

Sara

Resources that execute similar collections of activities are related. Sara is the only resource executing e and f . Therefore, she is not connected to other resources. Self-loops are suppressed as they contain no information (self-similarity)

Discovering organizational structures

PAGE 25

astart register

request

bexamine

thoroughly

cexamine casually

dcheck ticket

decide

pay compensation

reject request

reinitiate request

e

g

h

f

end

p3p1

p2 p4

p5

PeteMike Ellen

Sue SeanSara

Assistant

ManagerExpert

Another example

PAGE 26

a1

a2

a4

r1

process model

a3

a5

oe1

oe2 oe3

oe4 oe5

oe6 oe7 oe8

organizational model resources

r2

r3

r4

r5

r6

r7

r8

r9

Analyzing resource behavior, e.g., Yerkes-Dodson law of arousal

PAGE 27

Learning time and probabilities

• Replay, as before, but now considering timestamps.• Let us replay the first three cases in the event log:

− case 1 starts at time 12 and ends at time 54, − case 2 starts at time 17 and ends at time 73, − case 3 starts at time 25 and ends at time 98.

PAGE 28

PAGE 29

astart register

request

bexamine

thoroughly

cexamine casually

dcheck ticket

decide

pay compensation

reject request

reinitiate request

e

g

h

f

end

p3p1

p2 p4

p5

1,c:191,s:12

2,c:232,s:17

3,c:303,s:25

1,c:321,s:25

3,c:653,s:60

2,c:382,s:30

3,c:353,s:32

1,c:331,s:26

2,c:322,s:28

3,c:403,s:35

3,c:673,s:62

3,c:553,s:50

1,c:401,s:35

2,c:592,s:50

3,c:503,s:45

3,c:873,s:80

1,c:541,s:50

2,c:732,s:70

3,c:983,s:90

1:6

2:7

3:2

3:5

1:10

2:11

3:0

3:3

1:3

2:12

3:10

3:15

1:7

2:5

3:5

3:7

1:2

2:18

3:5

3:13

1:12

2:17

3:25

1:54

2:73

3:98

7

6

5

7

5

8

33

8

5

7

5

9

4

5

5

5

7

4

Another view on the timed replay of the first three cases

PAGE 30

abd

eh

0 10 20 30 40 50 60 70 80 90

ac

de

g

ab

cd d

e e

g

case 1

case 2

case 3

f

time

Timed replay projected onto resources

PAGE 31

a

b

d

e

h0 10 20 30 40 50 60 70 80 90

a

c

d

e

g

a

b

c

d

d

e e

g

f

Pete

Mike

EllenSue

Sean

Sara

time

Decision mining

PAGE 32

astart register

request

bexamine

thoroughly

cexamine casually

d

check ticket

decide

pay compensation

reject request

reinitiate request

e

g

h

f

end

c1

c2

c3

c4

c5

decision point #1

decision point #2

Example: XOR-split

PAGE 33

x

y

z

type region amount activity

gold south 987.30 z

type=gold and amount

Example: OR-split

PAGE 34

x

y

z

type region amount activity

gold south 987.30 y and z

type=gold or amount

Classification in process mining

• The application of classification techniques like decision tree learning is not limited to decision mining based on event/case data only.

• Additional predictor variables may be used:− behavioral information (count number of loops)− performance information (processing times)− contextual information (weather, queues, etc.)

• Alternative response variables can be analyzed:− uncover reasons for non-conformance (split

instances in two groups)− uncover reasons for delays

PAGE 35

Bringing it all together

PAGE 36

astart register

request

bexamine

thoroughly

cexamine casually

d

check ticket

decide

pay compensation

reject request

reinitiate request

e

g

h

f

end

c1

c2

c3

c4

c5

A

A

A

A

A

E

M

M

Pete

Mike

Ellen

Role A:Assistant

Sue

Sean

Role E:Expert

Sara

Role M:Manager

Step 2: create or discovera process model

Step 1: obtain an event log

Step 3: connect events in the log to activities in the

model

Step 4: extend the model

add

orga

niza

tiona

l pe

rspe

ctiv

e

add

time

pers

pect

ive

add

case

pers

pect

ive

add

othe

r pe

rspe

ctiv

es

Step 5: return integrated model

astart register

request

bexamine

thoroughly

cexamine casually

d

check ticket

decide

pay compensation

reject request

reinitiate request

e

g

h

f

end

c1

c2

c3

c4

c5

A

A

AM

eventlog

Chapter 8�Mining Additional PerspectivesOverviewMining additional perspectives�(one type of enhancement, cf. repair in context of conformance checking) Replay: Connecting events to model elements is essential for process miningRemember: Replay!Replay can detect problemsReplay can extract timing informationDecision mining: “Red” casesDecision mining: “Blue” casesStarting point: connected event log and modelProcessAttributes in event logsCases may also have attributesHelicopter view: Dotted charts Dotted chart for a process of a housing agency using absolute timeZooming inSame log, relative timeOrganizational miningResource-activity matrixSocial network analysisHandover of work matrixSocial network based on handover of work (threshold of 0.1)Handover of work at role levelProfileSocial network based on similarity of profilesDiscovering organizational structuresAnother exampleAnalyzing resource behavior, e.g., Yerkes-Dodson law of arousalLearning time and probabilitiesSlide Number 30Another view on the timed replay of the first three casesTimed replay projected onto resourcesDecision miningExample: XOR-splitExample: OR-split Classification in process miningBringing it all together