IDENTIFICATION METHOD OF POSSIBILITY DISTRIBUTIONS AND ITS APPLICATION TO DISCRIMINANT-ANALYSIS

This paper proposes an identification method of possibility distributi ons based on the data given by experts and leads to a possibility disc riminant rule from the viewpoint of possibility distributions. Since t he proposed method resorts to linear programming, possibility distribu tions defined as quadratic functions can easily be obtained. This meth od is used to construct if-then rules. The possibility discriminant ru le corresponds to Bayes' discriminant rule based on probability distri butions. To illustrate the proposed method, a discriminant problem is shown as a numerical example.