A pattern language approach to identify appropriate machine learning algorithms in the context of product development
Editor: Kevin Otto, Boris Eisenbart, Claudia Eckert, Benoit Eynard, Dieter Krause, Josef Oehmen, Nad
Author: Sonntag, Sebastian; Luttmer, Janosch; Pluhnau, Robin; Nagarajah, Arun
Institution: University of Duisburg-Essen
Section: Design Methods
DOI number: https://doi.org/10.1017/pds.2023.37
The product development process faces several challenges, such as an increasing and differentiated number of customer requirements, increasing product complexity, and shortened time-to-market. To address these challenges, the implementation of automation approaches in form of machine learning (ML) algorithms appears promising. However, companies lack the implementation of these approaches in their processes, inter alia due to inadequate knowledge and experience in this field. Therefore, the aim of this paper is to develop a structured formulized way of characterising ML algorithms, which can support non-experts in identifying the optimal algorithm to solve a given problem. First, existing approaches covering the determination of appropriate ML algorithms for a given task are examined. Based on this, a pattern language approach is introduced to characterise ML algorithms and problems, allowing matching to be performed to identify the most suitable one for a given task. Due to their broad application, the concept is demonstrated by creating patterns for decision trees and artificial neural networks. A study is conducted to prove that the proposed concept is appropriate to support the ML algorithm selection.