1996 Volume 8 Issue 6 Pages 1134-1143
When we wish to make a decision on the best approach to carry out a certain task at the present time, we usually refer to past cases whose conditions are similar to those of the present case. It is consequently useful to provide a decision maker with a system which retrieves the past cases. Since the completely same conditions of past cases to those of the present case do not exist, the system should evaluate a grade of the similarity and should output the retrieved cases accompanied with the grade. In the similar case retrieval, the cases are labeled with a number of indexes and they characterize the cases. The each index has a set of index attribute values and only one of them is designated to the corresponding index. A fuzzy thesaurus which shows quantitative relations of the index attribute values is, then, useful for the similar case retrieval. There is a conventional algorithm to generate a fuzzy thesaurus for document retrieval, so far, which is based on cooccurrences of a pair of keywords in the whole of the contents.It is, however, practical to generate the fuzzy thesaurus using only electronic data of the index. A problem of this case is that the cooccurrences in an index is not observed because there is only one index attribute value in each index.In this paper, a new algorithm to generate a fuzzy thesaurus using only the electronic data of index is presented and its application to the similar case retrieval is shown. This algorithm makes use of two indexes in a case and observes cross-occurrences of a pair of index attribute values in the two indexes. Using an operation, which we call sum-min composition, indirect relation of a pair of index attribute values in an index is evaluated based on the cross-occurrences and the fuzzy thesauri are generated.A decision support system to plan a rescue operation against a disaster has been developed. This system uses fuzzy thesauri to retrieve past rescue operations which are similar to the present condition. The presented algorithm has been applied to generate these fuzzy thesauri.
