Evaluating explicit functional relationships from fuzzy observations

Traditionally, when we investigate the dependence of some response variable on several regressors, we suppose that the values of the regressors are known exactly and random errors only occur with the response variable. In the present paper the author considers that both, the response variable and the regressors can only be known with inaccuracy, and he suggests a fuzzy approach as an alternative to the usual statistical treatment of the problem. This approach assumes that the response and the regressors are connected by a nonfuzzy explicit functional relationship involving an unknown nonfuzzy parameter, but the information available about this relationship may be described by means of several fuzzy observations. In this framework, two procedures (according to different purposes) are developed in order to transfer the information on the relationship to the parameter set. The first procedure is based on the expected cardinality, and the second one is based on the fuzzy expectation value. Some difficulties that can arise in connecting the last procedures with real problems, or in their numerical treatment, are then mentioned. Finally, the application of the suggested methods is illustrated by means of an example, in which the present approach is intuitively compared with the statistical one.

Recommendations

Cites work

Cited in

(24)

This page was built for publication: Evaluating explicit functional relationships from fuzzy observations

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1068497)