Lazy Conflict Detection with Genetic Algorithms

The customization of complex products and services requires configurators with often large and complex knowledge bases. In the case that configuration-related user requirements are inconsistent with the knowledge base, immediate feedback is desired. However, due to the domain’s complexity, efficient feedback generation is often not possible. In this paper we show how to use genetic algorithms to pre-generate minimal conflict sets. Their integration into the configurator allows response times required for interactive settings. Our evaluations, based on knowledge bases from the air pollution monitoring domain, show significant performance improvements.