Ontology-based Representation of Optimization Models

This article presents a new approach for representing and processing abstract optimization models. Confronted with model and data integration tasks for distributed Decision Support Systems which are especially composed out of software services, we describe model constituents such as constraints both structurally and semantically. Within our approach, typed model constituents can be integrated into complete models and the instantiation of model constituents itself with data and data models can be wrapped into semantic software services. Besides others, this supports the automated generation of adaptors and the search for and composition of services. The basic idea of our approach is to represent the optimization models as instance knowledge of different ontologies for both optimization and application domains. By separating the model expression structure from the goal and constraint conceptualizations predefined modeling constructs can be reused, where we do not only separate the model structure from data, but can also abstract the structure from a specific data model. We present an XML and ontology-query based approach for this separation and demonstrate the resulting flexible model integration procedure out of reusable goal and constraint types on a network flow problem.