Tolerating Inconsistency in Feature Models

Feature models have been widely adopted to reuse the requirements of a set of similar products in a domain. When constructing feature models, it is dicult to always ensure the consistency of feature models. Therefore, tolerating inconsistencies is important during the construction of feature models. The usual way of tolerating inconsistencies is to nd the minimal unsatisable core. However, identifying the minimal unsatisable core is time-consuming, which decreases itself the practicability. In this paper, we propose a priority based approach to tolerating inconsistencies in feature models eciently. The basic idea of our approach is to nd the weaker unsatised constraints, while keeping the rest of the feature model consistent. Our approach tolerates inconsistencies with the help of priority based operations while building feature models. To this end, we adopt the constraint hierarchy theory to express the degree of domain analysts’ condence on constraints (i.e. the priorities of constraints) and tolerate inconsistencies in feature models. Experiments have been conducted to demonstrate that our system can scale up to large feature models.