Development of a Knowledge-Based Model Management System

Abstract

This paper presents a software architecture and a graph-based framework for developing knowledge-based model management systems. The architecture consists of three major components: a model utilization subsystem, a modeling subsystem, and an inference engine. The core of the system is the inference engine that applies the graph-based framework to drive the processes of model integration and selection. The graph-based framework includes a model representation scheme and reasoning mechanisms. The representation scheme depicts a set of data as a node, a set of functions as an edge, and a basic model as a combination of two nodes and one connecting edge. Based on this scheme, mechanisms for model integration and selection are discussed. These mechanisms enable a model management system to create composite models automatically. A prototype implemented in PROLOG is also presented to demonstrate the graph-based framework.