A FRAMEWORK FOR JOIN PATTERN INDEXING IN INTELLIGENT DATABASE-SYSTEMS

In intelligent database systems, knowledge-directed inference often de rives large amounts of data, and the efficiency of query processing in these systems depends upon how the derived data are maintained. This paper focuses on situations where the rule is conditional on a join of multiple data objects (relations) and the rule-derived data are mater ialized to reduce the overall query processing:costs. We develop an in dexing technique based on a unique construct called join pattern relat ion, Several pattern redundancy reduction methods are also introduced to minimize the overhead cost of join indexing.