Abstract
Variability modeling allows to describe features, parameters and constraints in a product line. A common use of such model is to assist product configuration.Manually modeling variability being a long and tedious task, a growing number of techniques have been proposed to build feature models. Such techniques are generally specific to a given product line.
In this paper, we propose a more general modeling technique based on machine learning: we will train a classifier, using an oracle to predict the validity of samples. Our technique will thus require the making of such an oracle for each product line.
We will validate our method by evaluating its performances on two types of models: first we will use S.P.L.O.T. samples, which provide generic examples of feature models and second we will use OpenSCAD 3D objects available on Thingiverse as a more practical use case, where variability modeling was deemed necessary.
Date of Award | 27 Aug 2018 |
---|---|
Original language | French |
Awarding Institution |
|
Supervisor | Patrick Heymans (Supervisor) & MAXIME CORDY (Co-Supervisor) |
Keywords
- variability modeling
- feature model
- machine learning
- oracle
- S.P.L.O.T.
- 3D printing
- OpenSCAD
- Thingiverse