It is possible to classify the Object-Oriented design methods in two sets. The first set, which is the most numerous, uses an entity relationship approach to build a static class schema. Then this schema is completed by the use of different models (dynamic, functional). The second set tries to obtain a class schema but they don’t provide a guide to build it. So if different designers work on the same application, it is impossible to obtain the same schema.

So, in this paper we want to prove that we have defined a new design process (generalization process) which is based upon a set of rules to guide the users and the designers to build a representation of their application. The originality of our process lays upon works developped in machine learning and artificial intelligence. We try to translate the expertize of users or designers given through examples into a class schema. Therefore we have defined a new algorithm of clustering to organize examples into a hierarchy of classes.

This process is included in a project called C.O.D. (Class and Object Definition). The project is composed of different processes which take into account the expertize level of the designers and their knowledge about the application domain: specialization process, which is based upon generic application and fuzzy object classes; composition process, which uses a functional application definition with an algorithm to build classes and links between classes; generalization process, which is developed in this paper.

This content is only available via PDF.
You do not currently have access to this content.