Modeling by analogy : an approach to enhancing model management systems
Ting‐Peng Liang · Illinois Digital Environment for Access to Learning and Scholarship (University of Illinois at Urbana-Champaign) · 1989
Developing a productive and user-friendly modeling environment has been a focus in recent model management research.Onetechnique widely used in human modeling processes but virtually unexplored in existing literature is how analogical modeling can be supported by model management systems.Analogical modeling is a process by which model builders transform modeling knowledge obtained from previous experience to build models for new problems that share significant features with the previously solved ones.It reduces the need of repetitive trial and can increase the usefulness of model management systems.This paper investigates some fundamental issues of analogical modeling and how this technique can be supported by model management systems.First, models and related elements are clearly defined to avoid potential confusion resulting from different uses of the terms.Then, model similarity, the key to analogical modeling is discussed at the entity, structural, and functional levels of models.Finally, the analogical modeling process that allows new models to be developed by matching features with existing models and transforming functions accordingly is presented.An inventory control problem is used to illustrate how a new model can be developed analogically from an old one in order to meet the new inventory policy.