Home // ICSEA 2021, The Sixteenth International Conference on Software Engineering Advances // View article

SYS2VEC: System-to-Vector Latent Space Mappings

Authors:
Theo Mahmut Bulut
Vasil L. Tenev
Martin Becker

Keywords: system comparison; machine learning; graph similarity; product lines

Abstract:
As a product line evolves, new members emergeand existing ones are maintained – more or less in sync witheach other. In the context of long-living software and systemproduct lines, the capability to predict evolution trends withinthe structure of the various assets is essential. It helps tounderstand the underlying dependencies between work itemsnow and in the future and helps to make the product linearchitecture more robust against the predicted trends. Withthis, unnecessary erosion can be avoided and overall engineeringefficiency can be increased. With the increasing complexityof today’s systems, approaches that can identify and evaluatecommonalities, variabilities, and interdependencies in a largenumber of complex product variants and versions are gainingimportance. In order to increase efficiency, approaches thatsupport an incremental analysis in the space and time dimensionare desirable. A promising approach to this end is to map theversions of each variant to points in a vector space. Doing so,two challenges can be efficiently addressed: (i) the similaritymeasurement becomes the distance between vectors; and (ii)the estimation of evolutionary trends can be reduced to thewell-known interpolation problem. In this paper, we presentSYS2VEC – an approach for mapping product line variants intoa latent vector space by means of machine learning techniques.With our approach, we are able to show an increase in accuracyby a factor of 4 and halve the execution time compared to similarmachine-learning-based solutions.

Pages: 61 to 70

Copyright: Copyright (c) IARIA, 2021

Publication date: October 3, 2021

Published in: conference

ISSN: 2308-4235

ISBN: 978-1-61208-894-5

Location: Barcelona, Spain

Dates: from October 3, 2021 to October 7, 2021