Managing Derived Data in Intelligent Database Systems: An Implementation Study

Knowledge-directed inference in intelligent database systems derives large amounts of data that require new functionalities not available in conventional database management systems. This paper presents a front-end implementation of the selective materialization paradigm proposed in a previous study for efficient management of derived data. Empirical results obtained in this study demonstrate that selective materialization can be very cost effective compared to current approaches in other research prototypes. In addition, this paper introduces several concepts and algorithms key to the implementation, such as the minimal derivation graph, the rule transformation algorithm, and physical rule/data graphs.