Unsupervised Task Recognition from User Interaction Streams

Adrian Rebmann, Han van der Aa

Veröffentlichungen: Beitrag in BuchBeitrag in KonferenzbandPeer Reviewed

Abstract

User interaction events can give an accurate picture of tasks executed in a process, since they capture work performed across applications in a detailed manner. However, such data is too low level to be used for process analysis directly, since the underlying tasks are typically not apparent from individual events. Therefore, several task-recognition techniques were recently proposed that are able to abstract user interaction data to a higher level. However, these techniques work in an offline manner, requiring user interaction data to be stored in event logs. Such storage is often infeasible, though, due to the data’s sheer volume and its privacy-sensitive nature. While this can be avoided by analyzing user interaction data in a streaming manner, existing task-recognition techniques cannot be applied to such settings, since they require multiple, post-hoc passes over the entire data collection. To overcome this, we propose the first approach for unsupervised task recognition from user interaction streams. For a given stream, our approach continuously identifies task instances, groups them according to their type, and emits task-level events to an output stream. Our evaluation demonstrates our approach’s efficacy and shows that it outperforms two baseline approaches.

OriginalspracheEnglisch
TitelAdvanced Information Systems Engineering - 35th International Conference, CAiSE 2023, Proceedings
Redakteure*innenMarta Indulska, Iris Reinhartz-Berger, Carlos Cetina, Oscar Pastor
Herausgeber (Verlag)Springer Science and Business Media Deutschland GmbH
Seiten141-157
Seitenumfang17
ISBN (Print)9783031345593
DOIs
PublikationsstatusVeröffentlicht - 2023
Veranstaltung35th International Conference on Advanced Information Systems Engineering, CAiSE 2023 - Zaragoza, Spanien
Dauer: 12 Juni 202316 Juni 2023

Publikationsreihe

ReiheLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Band13901 LNCS
ISSN0302-9743

Konferenz

Konferenz35th International Conference on Advanced Information Systems Engineering, CAiSE 2023
Land/GebietSpanien
OrtZaragoza
Zeitraum12/06/2316/06/23

ÖFOS 2012

  • 102015 Informationssysteme
  • 102035 Data Science

Zitationsweisen