Titlebar

Export bibliographic data
Literature by the same author
plus on the publication server
plus at Google Scholar

 

Leveraging Small Sample Learning for Business Process Management

Title data

Käppel, Martin ; Schönig, Stefan ; Jablonski, Stefan:
Leveraging Small Sample Learning for Business Process Management.
In: Information and Software Technology. Vol. 132 (April 2021) . - Art.Nr. 106472.
ISSN 0950-5849
DOI: https://doi.org/10.1016/j.infsof.2020.106472

Abstract in another language

Context: Tool support for business process management (BPM) is improving more and more. Often, machine learning techniques are used to recognize certain execution patterns, to optimize workflows and to observe or predict processes. Frequently, many organisations cannot meet the fundamental prerequisites of machine learning methods since less data is recorded and therefore available for analysis. Most machine learning techniques rely on big and sufficient data source sets that can be analyzed. Small Sample Learning (SSL) tackles the issue of implementing machine learning methods in environments where only quantitatively insufficient datasets are available. These methods are strongly tailored to computer vision or natural language processing problems, which is why they are still neglected in the BPM area. Objective: This paper motivates the use of SSL methods in the BPM area and fosters a research stream that is concerned with the transferability to and the application of these methods in the BPM area. Method: We propose a concept for leveraging SSL methods in BPM and illustrate the idea exemplarly in the field of process mining. Results: Reasons for the need of SSL methods in the BPM area and a conceptual approach for transferring existing SSL methods to the BPM area. The feasibility of our apprach is shown by a brief overview of a primary study leveraging SSL methods for process prediction. Conclusions: Many areas of process mining or BPM in general depend on a sufficient amount of (training) data. Often small and medium sized companies lack ”big data”, which is why advantages of machine learning and data analysis in the context of BPM cannot be applied. Existing methods that deal with insufficient data are very domain-specific and must be transferred to the process mining area respectively the BPM area.

Further data

Item Type: Article in a journal
Refereed: Yes
Keywords: Small Sample Learning; BPM; Process Prediction; Machine Learning; Process Mining
Institutions of the University: Faculties > Faculty of Mathematics, Physics und Computer Science
Faculties > Faculty of Mathematics, Physics und Computer Science > Department of Computer Science
Faculties > Faculty of Mathematics, Physics und Computer Science > Department of Computer Science > Chair Applied Computer Science IV
Faculties > Faculty of Mathematics, Physics und Computer Science > Department of Computer Science > Chair Applied Computer Science IV > Chair Applied Computer Science IV - Univ.-Prof. Dr.-Ing. Stefan Jablonski
Result of work at the UBT: Yes
DDC Subjects: 000 Computer Science, information, general works > 004 Computer science
Date Deposited: 19 Apr 2021 09:36
Last Modified: 19 Apr 2021 09:36
URI: https://eref.uni-bayreuth.de/id/eprint/64805