A graph-theoretic approach to analyzing knowledge bases containing rules, models and data

Decision support systems have traditionally utilized stored data and decision models as sources of information. In recent years, such systems have also started to include a certain amount of expert knowledge, usually in the form of rules. Unfortunately, despite the evolution of systems containing all three types of resources, effective tools to comprehensively analyze the relationships between data relations, decision models and rules are still lacking. These relationships include the following: (1) a decision model may access a data relation to instantiate some required input, (2) a rule may define the circumstances under which a model is valid, (3) a rule may serve as a database integrity constraint, and (4) the antecedent of a rule may be instantiated through the execution of a decision model or retrieval of relevant data from a data relation. A numbver of approaches, including graph-theoretic approaches, have been used to analyze interactions among instances of each type of resource. However, these approaches generally are not effective for analyzing interactions such as those above, among heterogeneous components. In this paper, we show how metagraphs, a new graph-theoretic construct, can be used both to visualize knowledge base structure, as well as to analyze the various components to make useful inferences during problem solving.