Construction of linguistic variables from data and domain knowledge

This paper aims at construction of fuzzy linguistic variables by using rule induction and concept hierachy. Since representaiton of linguistic variables can be viewed as a linear combination of components, the algorithm is given as follows. First, domain knowledge (concept hiearchy) is used to transform an original dataset into transformed datasets. Contribution of a generalized concept to representation of an original variable is estimated by using indices obtained from transformed datasets.