Retrieving Cases from Relational Data-Bases: Another Stride Towards Corporate-Wide Case-Base Systems

Abstract

shimazu, akihiro joke.cl.nec.co.jp kitano spls26.ccs.mt.nec.co.jp Vital information for corporate activities is generally stored in large databases. While conventional data-base management systems offer limited query flexibility, systems capable of generating similarity-based queries, such as those seen in case-based reasoning research, would certainly enhance the utility of data resources. This paper describes a method for building case-based systems using a conventional relational data-base (RDB). The core of the algorithm is a novel approach to similarity computing in which database query form similarities, rather than similarities of individual cases, are computed. The method uses Standard Query Language (SQL) to achieve nearest neighbor matching, thus allowing similarity-based database retrieval. It has been implemented as a part of the CARET case retrieval tool and evaluated through the use of a newly developed corporate-wide case-based system for a software quality control domain. Experiments have shown the proposed method to provide retrieval results equivalent to those of non-RDB implementation at a sufficiently fast response time. 1