Weighing the Pros and Cons: Process Discovery with Negative Examples

Tijs Slaats; Søren Debois; Christoffer Olling Back

doi:10.1007/978-3-030-85469-0_6

Weighing the Pros and Cons: Process Discovery with Negative Examples

Tijs Slaats^*, Søren Debois, Christoffer Olling Back

^*Corresponding author for this work

Software, Data, People & Society

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

8 Citations (Scopus)

Abstract

Contemporary process discovery methods take as inputs only positive examples of process executions, and so they are one-class classification algorithms. However, we have found negative examples to also be available in industry, hence we propose to treat process discovery as a binary classification problem. This approach opens the door to many well-established methods and metrics from machine learning, in particular to improve the distinction between what should and should not be allowed by the output model. Concretely, we (1) present a formalisation of process discovery as a binary classification problem; (2) provide cases with negative examples from industry, including real-life logs; (3) propose the Rejection Miner binary classification procedure, applicable to any process notation that has a suitable syntactic composition operator; and (4) apply this miner to the real world logs obtained from our industry partner, showing increased output model quality in terms of accuracy and model size.

Original language	English
Title of host publication	Business Process Management - 19th International Conference, BPM 2021, Proceedings
Editors	Artem Polyvyanyy, Moe Thandar Wynn, Amy Van Looy, Manfred Reichert
Publisher	Springer
Publication date	2021
Pages	47-64
ISBN (Print)	9783030854683
DOIs	https://doi.org/10.1007/978-3-030-85469-0_6
Publication status	Published - 2021
Event	19th International Conference on Business Process Management, BPM 2021 - Rome, Italy Duration: 6 Sep 2021 → 10 Sep 2021

Conference

Conference	19th International Conference on Business Process Management, BPM 2021
Country/Territory	Italy
City	Rome
Period	06/09/2021 → 10/09/2021

Series	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	12875 LNCS
ISSN	0302-9743

Bibliographical note

Publisher Copyright:
© 2021, Springer Nature Switzerland AG.

Keywords

Binary classification
Labelled event logs
Negative examples
Process mining

Access to Document

10.1007/978-3-030-85469-0_6

Cite this

Slaats, T., Debois, S., & Back, C. O. (2021). Weighing the Pros and Cons: Process Discovery with Negative Examples. In A. Polyvyanyy, M. T. Wynn, A. Van Looy, & M. Reichert (Eds.), Business Process Management - 19th International Conference, BPM 2021, Proceedings (pp. 47-64). Springer. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) Vol. 12875 LNCS https://doi.org/10.1007/978-3-030-85469-0_6

Weighing the Pros and Cons : Process Discovery with Negative Examples. / Slaats, Tijs; Debois, Søren; Back, Christoffer Olling.

Business Process Management - 19th International Conference, BPM 2021, Proceedings. ed. / Artem Polyvyanyy; Moe Thandar Wynn; Amy Van Looy; Manfred Reichert. Springer, 2021. p. 47-64 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 12875 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

Slaats, T, Debois, S & Back, CO 2021, Weighing the Pros and Cons: Process Discovery with Negative Examples. in A Polyvyanyy, MT Wynn, A Van Looy & M Reichert (eds), Business Process Management - 19th International Conference, BPM 2021, Proceedings. Springer, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 12875 LNCS, pp. 47-64, 19th International Conference on Business Process Management, BPM 2021, Rome, Italy, 06/09/2021. https://doi.org/10.1007/978-3-030-85469-0_6

Slaats T, Debois S, Back CO. Weighing the Pros and Cons: Process Discovery with Negative Examples. In Polyvyanyy A, Wynn MT, Van Looy A, Reichert M, editors, Business Process Management - 19th International Conference, BPM 2021, Proceedings. Springer. 2021. p. 47-64. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 12875 LNCS). https://doi.org/10.1007/978-3-030-85469-0_6

Slaats, Tijs ; Debois, Søren ; Back, Christoffer Olling. / Weighing the Pros and Cons : Process Discovery with Negative Examples. Business Process Management - 19th International Conference, BPM 2021, Proceedings. editor / Artem Polyvyanyy ; Moe Thandar Wynn ; Amy Van Looy ; Manfred Reichert. Springer, 2021. pp. 47-64 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 12875 LNCS).

@inproceedings{3b5c696a99504dca970c20b46ce16551,

title = "Weighing the Pros and Cons: Process Discovery with Negative Examples",

abstract = "Contemporary process discovery methods take as inputs only positive examples of process executions, and so they are one-class classification algorithms. However, we have found negative examples to also be available in industry, hence we propose to treat process discovery as a binary classification problem. This approach opens the door to many well-established methods and metrics from machine learning, in particular to improve the distinction between what should and should not be allowed by the output model. Concretely, we (1) present a formalisation of process discovery as a binary classification problem; (2) provide cases with negative examples from industry, including real-life logs; (3) propose the Rejection Miner binary classification procedure, applicable to any process notation that has a suitable syntactic composition operator; and (4) apply this miner to the real world logs obtained from our industry partner, showing increased output model quality in terms of accuracy and model size.",

keywords = "Binary classification, Labelled event logs, Negative examples, Process mining",

author = "Tijs Slaats and S{\o}ren Debois and Back, {Christoffer Olling}",

note = "Publisher Copyright: {\textcopyright} 2021, Springer Nature Switzerland AG.; 19th International Conference on Business Process Management, BPM 2021 ; Conference date: 06-09-2021 Through 10-09-2021",

year = "2021",

doi = "10.1007/978-3-030-85469-0_6",

language = "English",

isbn = "9783030854683",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer",

pages = "47--64",

editor = "Artem Polyvyanyy and Wynn, {Moe Thandar} and {Van Looy}, Amy and Manfred Reichert",

booktitle = "Business Process Management - 19th International Conference, BPM 2021, Proceedings",

address = "Switzerland",

}

TY - GEN

T1 - Weighing the Pros and Cons

T2 - 19th International Conference on Business Process Management, BPM 2021

AU - Slaats, Tijs

AU - Debois, Søren

AU - Back, Christoffer Olling

PY - 2021

Y1 - 2021

N2 - Contemporary process discovery methods take as inputs only positive examples of process executions, and so they are one-class classification algorithms. However, we have found negative examples to also be available in industry, hence we propose to treat process discovery as a binary classification problem. This approach opens the door to many well-established methods and metrics from machine learning, in particular to improve the distinction between what should and should not be allowed by the output model. Concretely, we (1) present a formalisation of process discovery as a binary classification problem; (2) provide cases with negative examples from industry, including real-life logs; (3) propose the Rejection Miner binary classification procedure, applicable to any process notation that has a suitable syntactic composition operator; and (4) apply this miner to the real world logs obtained from our industry partner, showing increased output model quality in terms of accuracy and model size.

AB - Contemporary process discovery methods take as inputs only positive examples of process executions, and so they are one-class classification algorithms. However, we have found negative examples to also be available in industry, hence we propose to treat process discovery as a binary classification problem. This approach opens the door to many well-established methods and metrics from machine learning, in particular to improve the distinction between what should and should not be allowed by the output model. Concretely, we (1) present a formalisation of process discovery as a binary classification problem; (2) provide cases with negative examples from industry, including real-life logs; (3) propose the Rejection Miner binary classification procedure, applicable to any process notation that has a suitable syntactic composition operator; and (4) apply this miner to the real world logs obtained from our industry partner, showing increased output model quality in terms of accuracy and model size.

KW - Binary classification

KW - Labelled event logs

KW - Negative examples

KW - Process mining

U2 - 10.1007/978-3-030-85469-0_6

DO - 10.1007/978-3-030-85469-0_6

M3 - Article in proceedings

AN - SCOPUS:85115196600

SN - 9783030854683

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 47

EP - 64

BT - Business Process Management - 19th International Conference, BPM 2021, Proceedings

A2 - Polyvyanyy, Artem

A2 - Wynn, Moe Thandar

A2 - Van Looy, Amy

A2 - Reichert, Manfred

PB - Springer

Y2 - 6 September 2021 through 10 September 2021

ER -