A fuzzy approach to case retrieval through fuzzy extension of SQL

The use of database technologies for implementing case-based reasoning techniques is attracting a lot of attention for several reasons. First, the possibility of using standard DBMS for storing and representing cases significantly reduces the effort needed to develop a CBR system; in fact, data of interest are usually already stored into relational databases and used for different purposes as well. Finally, the use of standard query languages, like SQL, may facilitate the introduction of a case-based system into the real-world, by putting retrieval on the same ground of normal database queries. Unfortunately, SQL is not able to deal with queries like those needed in a CBR system, so different approaches have been tried, in order to build retrieval engines able to exploit, at the lower level, standard SQL. In this paper, we concentrate on Fuzzy Case Retrieval where case similarity is defined by means of fuzzy predicates, operators and standard fuzzy logic connectives, rather than through distance measures as in usual k-NN approaches. We present a proposal where case retrieval is implemented by using a straightforward fuzzy extension to standard SQL, where the boolean satisfiability condition for tuple selection is substituted with a fuzzy one. A case-based client/server architecture exploiting Fuzzy-SQL as a retrieval engine is then presented, together with some possible applications in e-commerce and medical domains.