Definition of a similarity measure between cases based on auto/cross-fuzzy thesauri

✍ Scribed by Kohei Nomoto; Wakasa Kise; Yoshio Kosuge

• Publisher: John Wiley and Sons
• Year: 2002
• Tongue: English
• Weight: 163 KB
• Volume: 33
• Category: Article
• ISSN: 0882-1666
• DOI: 10.1002/scj.1156

✦ Synopsis

Abstract

A similarity measure between cases is needed in order to evaluate the degree of similarity when using past similar cases in order to resolve current problems. In similar case retrieval, multiple indices are set up in order to characterize the queries and individual cases, then terms are given as values to each. The similarity measure between cases commonly used is defined using the rate at which the values provided from the corresponding indices match. In practice, however, values cannot be expected to be mutually exclusive. As a result, a natural expansion of this approach is to have relationships in which mutually similar meanings are reflected in the similarity measure between cases. In this paper the authors consider an auto‐fuzzy thesaurus which gives the relationship for values between corresponding indices and a cross‐fuzzy thesaurus which gives the relationship for values between mutually distinct indices, then defines a similarity measure between cases which considers the relationship of index values based on these thesauri. This definition satisfies the characteristics required for the operation of case‐based retrieval even when one value is not necessarily given in the index. Finally, using a test similar case retrieval system, the authors perform a comparative analysis of the proposed similarity measure between cases and a conventional approach. © 2002 Wiley Periodicals, Inc. Syst Comp Jpn, 33(9): 99–108, 2002; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/scj.1156