Combining a knowledge-based system and a clustering method for a construction of models in ill-structured domains

Standard statistical methods usually ignore the additional information that an expert has about the domain structure. Direct treatment of symbolic information is not a very common characteristic of statistical systems. KLASS is a statistical clustering system that provides the possibility of using either quantitative and qualitative variables in the domain description. The user may also declare part of its knowledge about the domain structure. The system is especially useful when dealing with ill-structured domains (i.e. a domain where the consensus among the experts is weak as mental diseases, sea sponges, books, painters…). That is why it is also useful from the artificial intelligence point of view. The output is a partition of the target domain. Conceptual and extensional descriptions of the classes can also be achieved.