On the design of a classifier with linguistic variables as inputs (Q793514)

In this paper a problem of classification of patterns given in the linguistic form is investigated. Both input and output variables are realized in the form of fuzzy subsets of the universe set. A classifier is then designed as a fuzzy relation from the set of input names of patterns to the set of output names of classes. In the first stage of the development of the classifier a starting approximation is formed. This relation contains an expert's knowledge about the nature of the problem. In the second stage, the classifier obtained from the answers of the expert can be improved using appropriate sequences \(\{X_ k\}\) and \(\{Y_ k\}\) where \(X_ k\) and \(Y_ k\) are an input fuzzy subset of patterns and a fuzzy output subset of classes, respectively. An iterative algorithm for refinement of the classifier is suggested and illustrated by an example.