A fuzzy approach to similarity in Case-Based Reasoning suitable to SQL implementation

The aim of this paper is to formally introduce a notion of acceptance and similarity, based on fuzzy logic, among case features in a case retrieval system. This is pursued by rst reviewing the relationships between distance-based similarity (i.e. the standard approach in CBR) and fuzzy-based similarity, with particular attention to the formalization of a case retrieval process based on fuzzy query specication. In particular, we present an approach where local acceptance relative to a feature can be expressed through fuzzy distributions on its domain, abstracting the actual values to linguistic terms. Furthermore, global acceptance is completely grounded on fuzzy logic, by means of the usual combinations of local distributions through specic dened norms. We propose a retrieval architecture, based on the above notions and realized through a fuzzy extension of SQL, directly implemented on a standard relational DBMS. The advantage of this approach is that the whole power of an SQL engine can be fully exploited, with no need of implementing specic retrieval algorithms. The approach is illustrated by means of some examples from a recommender system called MyWine, aimed at recommending the suitable wine bottles to a customer providing her requirements in both crisp and fuzzy way.