An object-oriented approach for integrating data and model bases in decision support systems

Decision support systems (DSS) have been developed for many different applications over the last two decades. The individual components of DSS--the data management system, the model management system and the user interface--have been topics of research for several years. Often a mix of data and models may be needed to generate solutions to problems requiring computer-based support. This dissertation develops a common approach for representing data and models in a semantically expressive form and for manipulating DSS components in a seamless manner.;In the approach developed in this dissertation, a query initiated by a user is answered with database retrieval or by model selection and execution (which would also require database accesses). The query solution path and processing are transparent to the user. The query may be answered through one or more database accesses or by a manipulation of the data (i.e. by solving a model) to derive an answer to the user's query. The data-model integration process for query solution is based on a four-level, object-oriented framework for DSS development. The framework postulates that DSS development takes place in four stages: object identification--a top-down approach for studying the domain to identify objects; object specification--a bottom-up approach for analyzing user views; object class structuring-- defining relationships between objects within a class; object class linking--illustrating the seamless use of data and models by specifying three static and dynamic links between data and models.;An advantage of the four-stage object-oriented framework for DSS development is that both data and models can be represented using the same paradigm. Object-oriented constructs are also used in manipulating data and models. Finally, the use of an object-oriented model for representing the components of a DSS has the added advantage of being extensible, in that new classes of data and models can be added at any time.;The feasibility of the framework was tested by implementing a prototype DSS with data and models drawn from the production planning and control (PPC) domain. A conceptual model for the domain was developed based on the four-stage object-oriented framework for DSS development. The prototype was implemented using the SMALLTALK object-oriented programming environment on an IBM PS/2. The framework was successfully validated by testing the prototype DSS with several queries, involving complex manipulations of data and models.