From Low-Level Events to Activities - Process Mining · From Low-Level Events to Activities A Pattern-based Approach Felix Mannhardt, Massimiliano de Leoni, Hajo A. Reijers, Wil M.P

From Low-Level Events to ActivitiesA Pattern-based Approach

Felix Mannhardt, Massimiliano de Leoni, Hajo A. Reijers,Wil M.P. van der Aalst, Pieter J. Toussaint (NTNU, Trondheim)

Paper: 10.1007/978-3-319-45348-4_8

Scope: Analysis of event data

PAGE 1

Problem: Events ≠ Recognizable activities

PAGE 2

Alarm

records

Event Time

Red Button 20:08:00

Green Button 20:10:00

Gray Button 20:16:00

Handover

Event Time

Nurse Changed 07:02:00



records

Event Time



records

Visit

Goal 1: From low-level to high-level events (Supervised)

PAGE 3

Event Log Abstracted Log

Low-level Event Time

Red Button 20:08:00








High-level Event Start Complete

Alarm 20:08:00 20:16:00

Visit 22:02:00 22:10:00

Shift Change 07:02:00 07:18:00

Goal 2: Deal with shared labels, concurrency and noise

PAGE 4


Red Button 20:08:00


Gray Button 20:16:00Green Button 22:02:00

Nurse Changed 22:09:00Gray Button 22:10:00

Nurse Changed 22:15:00Nurse Changed 07:02:00



High-level Event Start Complete

Alarm 20:08:00 20:10:00

Visit 22:02:00 22:10:00

Handover 07:02:00 07:18:00

Missing events

Unexpected events

Event Log Abstracted Log

Related work

• Unsupervised event abstraction (Ferreira et al., Folino et al., …)• Does not take domain knowledge into account

• Supervised event abstraction (Thomas Baier et al., …)• Assumes knowledge of a complete process model• Semi-automatic discovery of the mapping between events and activities• Uses clustering and constraint satisfaction to determine the mapping

• Complex event processing• Focus on detection of event patterns in data streams• No concept of process instance / trace

PAGE 5

Overview: From Low-level Events to Activities

PAGE 6

Event Log Aligned Log Abstracted Log

Activity Pattern

Abstraction Model

3) Align model & log 4) Abstract events

1) Encode knowledge 2) Compose model

1) Encode knowledge on activities as activity patterns

PAGE 7

Alarm Handover

GreenA

GrayA

RedA

TA

TA'-TA ≤ 10GreenS

GrayS

Nurse S

TS

TS'-TS ≤ 30

GreenV

GrayV

Pattern defines traces expected for one activity instance!

Data Petri net used for clear semantics!

Visit

What about interaction between activity patterns?

PAGE 8

Alarm

Interaction?GreenA

GrayA

RedA

TA

TA'-TA ≤ 10

GreenV

GrayV

Visit

From Low-level Events to Activities

PAGE 9


Activity Pattern

Abstraction Model

2) Compose model

2) Build an integrated abstraction model

PAGE 10

Interaction?Alarm Visit

Parallel

Alarm Visit

Interleaving

↔

Alarm Visit

Choice

✖

Alarm Visit

Sequence

Alarm Visit Alarm0..✱

Repetition

GreenV

GrayVGreenS

GrayS

Nurse S

TSTS'-TS ≤ 30

GreenA

GrayA

RedA

TA

TA'-TA ≤ 10 `

2) Build an integrated abstraction model

PAGE 11

Handover

Alarm Visit

↔

0..✱

0..✱

0..✱

0..✱

Automaticcompilation

Abstraction Model

Compiled Abstraction Model (DPN)


PAGE 12


Activity Pattern

Abstraction Model

3) Align model & log

3) Align event log to abstraction model

PAGE 13

Event Log


Red Button 20:08:00








G B tt 07 18 00

GreenV

GrayVGreenS

GrayS

Nurse S

TSTS'-TS ≤ 30

GreenA

GrayA

RedA

TA

TA'-TA ≤ 10 `

Compiled abstraction model

Alignment


PAGE 14


Red Button 20:08:00








G B tt 07 18 00

Process Step Time

RedA 20:08:00

GreenA 20:10:00

GrayA

GreenV 22:02:00

GrayV 22:10:00

NurseS 07:02:00

GreenS 07:15:00

G 07 18 00

Event Log Existing alignment methods Aligned Log

Model only

Log only

…


PAGE 15


Activity Pattern

Abstraction Model

4) Abstract events

4) Create the abstracted event log with high-level events

PAGE 16

Aligned LogAbstract aligned events

Abstracted Log

High-level Event Transition Time

Alarm start 20:08:00

Alarm complete 20:10:00

Visit start 22:02:00

Visit complete 22:10:00

Handover start 07:02:00

Handover complete 07:18:00

Process Step Time

RedA 20:08:00

GreenA 20:10:00

GrayA

GreenV 22:02:00

GrayV 22:10:00

NurseS 07:02:00

GreenS 07:15:00

GrayS 07:18:00

Alarm

Visit

Handover

Use time of previous mapped event.

Recap: Abstraction Method

PAGE 17


Activity Pattern

Abstraction Model


4) Abstract based on aligned log

1) Encode knowledgeon activities as patterns

2) Build integrated model

Evaluation: Digital whiteboard system in a hospital

PAGE 18

• Information system• Digital whiteboard• Supports work of nurses• Mixed clinical & logistic info• Flexible system

• Dataset• One year• > 8,000 cases• > 280,000 events• Event per changed cell

Evaluation: Activity Patterns

PAGE 19

• 18 activity patterns• Our assumptions• Interview with expert

• Abstraction model• Most interleaved & repeated• Five concurrent activities

• Resulting abstraction• Low average error rate• Successful abstraction

Evaluation: Detected shift change pattern

PAGE 20

Shift pattern captures a meaningful activity

Blue: Nurse ChangedGreen: Call Signal GreenYellow: Call Signal Gray Relatively rare pattern!

00:00 24:00 00:00 24:00

Conclusion & Future work

PAGE 21

• Method• Pattern-based event abstraction• Knowledge encoded as activity patterns• Abstraction using alignment methods

• Results• Handles shared labels, concurrency and noise• Alignment gives reliability measure • Successfully used in a case study

• Future work• In-depth comparison to related methods• Prioritization among activity patterns• Decomposed of alignment methods Implemented in ProM 6.6

Questions?

PAGE 22

@fmannhardt - [email protected]

http://promtools.org - LogEnhancement package

http://promtools.org/

Documents

From Low-Level Events to Activities - Process Mining · From Low-Level Events to Activities A Pattern-based Approach Felix Mannhardt, Massimiliano de Leoni, Hajo A. Reijers, Wil M.P