ARTIFICIAL NEURAL NETWORK AND ENTROPY APPROACH IN FUZZY NONLINEAR REGRESSION

Abstract

Fuzzy nonlinear regression (FNR) is different from classic regression models just because its output consists of fuzzy numbers. Predictions are realized by FNR models  for the cases in which both input variables are nonlinearly related and output variable is fuzzy.  Besides, a FNR model may be used to construct a probability interval for the output variable precisely. It is important to note that an entropy-based approach to FNR models results in smaller propagations for fuzzy intervals.