Process-based reconstructive approach to model building

Abstract

The role of derivational analogy approach in the model building domain is investigated. In particular, the paper focuses on enhancing decision support systems with model construction capabilities so that the prior modeling experience can be conveyed into new problem situations. The role of analogy in the model construction process is pointed out. This is followed by a discussion of the different approaches to analogical reasoning and the assessment of their suitability for model building decision support systems. The proposed approach, process analogies, is detailed as to required knowledge sources and representation schemes for problem and model description, and illustrated using examples. The reusable model pieces in the linear programming modeling domain are identified at various levels of abstraction and their impact on model base organization is pointed out. The alternative model base organization strategies are contrasted in terms of storage, search and reasoning tradeoffs.