Download - Measuring the Precision of Multi-perspective Process Models

Transcript
Page 1: Measuring the Precision of Multi-perspective Process Models

Measuring the Precision ofMulti-perspective Process Models

Felix Mannhardt joint work withMassimiliano de Leoni, Hajo A. Reijers,Wil M.P. van der Aalst

Page 2: Measuring the Precision of Multi-perspective Process Models

Department of Mathematics and Computer Science

Precision

PAGE 2 / 8

β€œFlower Model” lacking any precision

B

C

A

Page 3: Measuring the Precision of Multi-perspective Process Models

Department of Mathematics and Computer Science

Precision of Multi-perspective Process Models

SimpleCheck

ExtensiveCheck

Handle Request Decide

Loan

PAGE 3 / 8

A π‘π‘Ÿπ‘’π‘π‘–π‘ π‘–π‘œπ‘›π΄

SimpleCheck

ExtensiveCheck

Handle Request Decide

Loan

Loan < 2,000

Loan > 1,000

Bπ‘π‘Ÿπ‘’π‘π‘–π‘ π‘–π‘œπ‘›π΅

π‘π‘Ÿπ‘’π‘π‘–π‘ π‘–π‘œπ‘›π΅>π‘π‘Ÿπ‘’π‘π‘–π‘ π‘–π‘œπ‘›π΄

Existing work ignores added precision by multi-perspective rules / constraints

Page 4: Measuring the Precision of Multi-perspective Process Models

Department of Mathematics and Computer Science

Approach: Multi-perspective Precision

PAGE 4 / 8

Multi-perspective Process Model (P)

Fitting EventLog (L)

Precision[0..1]

INPUT OUTPUT

APPROACH

βˆ‘π’†βˆˆπ‘³

𝒐𝒃𝒔 𝒆𝒓𝒗𝒆𝒅𝑷 (𝒆)π’‘π’“π’†π’„π’Šπ’”π’Šπ’π’ (𝑷 ,𝑳)=ΒΏ βˆ‘

π’†βˆˆπ‘³π’‘π’π’”π’”π’Šπ’ƒπ’π’†π‘· (𝒆)

Page 5: Measuring the Precision of Multi-perspective Process Models

Department of Mathematics and Computer Science

SimpleCheck

ExtensiveCheck

Handle Request Decide

Loan

Loan < 2,000

Loan > 1,000

SimpleCheck

ExtensiveCheck

Handle Request Decide

Loan

Loan < 2,000

Loan > 1,000

Precision: Observed / Possible Behavior

PAGE 5 / 8

SimpleCheck

ExtensiveCheck

Handle Request Decide

Loan

Loan < 2,000

Loan > 1,000C Id Event Loan obs pos1 Handle Request 800

1 Simple Check -

1 Decide -

2 Handle Request 1800

2 Ext. Check -

2 Decide -

SimpleCheck

ExtensiveCheck

Handle Request Decide

Loan

Loan < 2,000

Loan > 1,000

𝒐𝒃𝒔 𝑷 (π’†πŸ )=|{Handle Request }|=πŸπ’‘π’π’” 𝑷 (π’†πŸ )=|{Handle Request }|=𝟏

𝒔𝒕𝒂𝒕𝒆 (π’†πŸ )=(ΒΏ>, {})

C Id Event Loan obs pos1 Handle Request 800 1 11 Simple Check -

1 Decide -

2 Handle Request 1800

2 Ext. Check -

2 Decide -

C Id Event Loan obs pos1 Handle Request 800 1 1

1 Simple Check - 1 11 Decide -

2 Handle Request 1800

2 Ext. Check -

2 Decide -𝒐𝒃𝒔 𝑷 (π’†πŸ )=|{Simple }̌|=πŸπ’‘π’π’” 𝑷 (π’†πŸ )=ΒΏ

𝒔𝒕𝒂𝒕𝒆 (π’†πŸ )=(¿𝐻>, {𝐿=800 })

SimpleCheck

ExtensiveCheck

Handle Request Decide

Loan

Loan < 2,000

Loan > 1,000C Id Event Loan obs pos1 Handle Request 800 1 1

1 Simple Check - 1 1

1 Decide - 1 12 Handle Request 1800

2 Ext. Check -

2 Decide -𝒐𝒃𝒔 𝑷 (π’†πŸ‘ )=|{Decide }|=πŸπ’‘π’π’” 𝑷 (π’†πŸ‘ )=|{Decide }|=𝟏

𝒔𝒕𝒂𝒕𝒆 (π’†πŸ‘ )=(¿𝐻 ,𝑆>, {𝐿=800 })

SimpleCheck

ExtensiveCheck

Handle Request Decide

Loan

Loan < 2,000

Loan > 1,000C Id Event Loan obs pos1 Handle Request 800 1 1

1 Simple Check - 1 1

1 Decide - 1 1

2 Handle Request 1800 1 12 Ext. Check -

2 Decide -𝒐𝒃𝒔 𝑷 (π’†πŸ’ )=|{Handle Request }|=πŸπ’‘π’π’” 𝑷 (π’†πŸ’ )=|{Handle Request }|=𝟏

𝒔𝒕𝒂𝒕𝒆 (π’†πŸ’ )=(ΒΏ , {})

SimpleCheck

ExtensiveCheck

Handle Request Decide

Loan

Loan < 2,000

Loan > 1,000C Id Event Loan obs pos1 Handle Request 800 1 1

1 Simple Check - 1 1

1 Decide - 1 1

2 Handle Request 1800 1 1

2 Ext. Check - 1 22 Decide -𝒐𝒃𝒔 𝑷 (π’†πŸ“ )=|{Ext . }̌|=𝟏

𝒑𝒐𝒔 𝑷 (π’†πŸ“ )=|{Ext .Ceπ‘π‘˜ ,π‘†π‘–π‘šπ‘π‘™π‘’ h𝐢 π‘’π‘π‘˜}|=𝟐

𝒔𝒕𝒂𝒕𝒆 (π’†πŸ“ )=(¿𝐻>, {𝐿=1800 })

C Id Event Loan1 Handle Request 800 1 1

1 Simple Check - 1 1

1 Decide - 1 1

2 Handle Request 1800 1 1

2 Ext. Check - 1 2

2 Decide - 1 1 𝒑𝒐𝒔 𝑷 (π’†πŸ” )=|{𝐷𝑒𝑐𝑖𝑑𝑒 }|=𝟏

𝒔𝒕𝒂𝒕𝒆 (π’†πŸ” )=(¿𝐻 ,𝐸>, {𝐿=1800 })

SimpleCheck

ExtensiveCheck

Handle Request Decide

Loan

Loan < 2,000

Loan > 1,000C Id Event Loan1 Handle Request 800 1 1

1 Simple Check - 1 1

1 Decide - 1 1

2 Handle Request 1800 1 1

2 Ext. Check - 1 2

2 Decide - 1 1

6 7

𝒐𝒃𝒔 𝑷 (𝒆 )=|{observed activities at state }|𝒑𝒐𝒔 𝑷 (𝒆 )=ΒΏ {π‘π‘œπ‘ π‘ π‘–π‘π‘™π‘’π‘Žπ‘π‘‘π‘–π‘£π‘–π‘‘π‘–π‘’π‘  π‘Žπ‘‘ π‘ π‘‘π‘Žπ‘‘π‘’ }∨¿

𝒔𝒕𝒂𝒕𝒆 (𝒆 )=π‘ π‘‘π‘Žπ‘‘π‘’π‘œπ‘“ h𝑑 π‘’π‘π‘Ÿπ‘œπ‘π‘’π‘ π‘ π‘šπ‘œπ‘‘π‘’π‘™

Page 6: Measuring the Precision of Multi-perspective Process Models

Department of Mathematics and Computer Science

Full Example for Model A & Model B

PAGE 6 / 8

SimpleCheck

ExtensiveCheck

Handle Request Decide

Loan

AC Id Event Loan1 Handle Request 800 1 1

1 Simple Check - 1 2

1 Decide - 1 1

2 Handle Request 1800 1 1

2 Extensive Check - 2 2

2 Decide - 1 1

3 Handle Request 1800 1 1

3 Simple Check - 2 2

3 Decide - 1 1

4 Handle Request 2500 1 1

4 Extensive Check - 1 2

4 Decide - 1 1

14 16

Model A

π‘π‘Ÿπ‘’π‘π‘–π‘ π‘–π‘œπ‘› (𝑃 ,𝐿 )=βˆ‘π‘’βˆˆ 𝐿

π‘œπ‘π‘ π‘’π‘Ÿπ‘£π‘’π‘‘π‘ƒ(𝑒)=14  

βˆ‘π‘’βˆˆπΏ

π‘π‘œπ‘ π‘ π‘–π‘π‘™π‘’π‘ƒ (𝑒)=16  β‰ˆπŸŽ .πŸ–πŸ•πŸ“

SimpleCheck

ExtensiveCheck

Handle Request Decide

Loan

Loan < 2,000

Loan > 1,000B

C Id Event Loan1 Handle Request 800 1 1

1 Simple Check - 1 11 Decide - 1 1

2 Handle Request 1800 1 1

2 Extensive Check - 2 2

2 Decide - 1 1

3 Handle Request 1800 1 1

3 Simple Check - 2 2

3 Decide - 1 1

4 Handle Request 2500 1 1

4 Extensive Check - 1 14 Decide - 1 1

14 14

Model B

π‘π‘Ÿπ‘’π‘π‘–π‘ π‘–π‘œπ‘› (𝑃 ,𝐿 )=βˆ‘π‘’βˆˆ 𝐿

π‘œπ‘π‘ π‘’π‘Ÿπ‘£π‘’π‘‘π‘ƒ(𝑒)=14  

βˆ‘π‘’βˆˆ 𝐿

π‘π‘œπ‘ π‘ π‘–π‘π‘™π‘’π‘ƒ(𝑒)=14  β‰ˆπŸ

Page 7: Measuring the Precision of Multi-perspective Process Models

Evaluation on Road Fines Log

Inductive Miner Inductive Miner & Rules

Normative Model Normative Model & Rules

0.00.10.20.30.40.50.60.70.80.91.0

0.300.36

0.64

0.83

ETC Precision Precision Fitness

Page 8: Measuring the Precision of Multi-perspective Process Models

Department of Mathematics and Computer Science

Summary

PAGE 8 / 8

β€’ 1st precision measure for multi-perspective process modelsβˆ’ Fast to calculateβˆ’ Flexible frameworkβˆ’ Implemented in ProM

β€’ Preliminary Evaluationβˆ’ Illustrative examplesβˆ’ Real-life dataset with > 500,000 events

Handle750 Simple Decide

Handle1250 Ext. Decide

Handle5000 Simple Decide

Handle

750Simple Decide

Handle

1500Simpl

eDecid

e

SimpleCheck

ExtensiveCheck

Handle Request Decide

Loan

Loan < 2,000

Loan > 1,000

precision [0..1]

Page 9: Measuring the Precision of Multi-perspective Process Models

Department of Mathematics and Computer ScienceImage source: http://commons.wikimedia.org/wiki/File:Pictofigo_-_Idea.png

Questions? Remarks? Ideas?