Numerical solution of fuzzy interval-valued regression model by neural network

In this paper, a fuzzy linear regression model with triangular interval-valued fuzzy output and inputs is presented. The presented mathematical formulation of the model leads to a non-differentiable program which is converted to an equivalent convex problem by using an appropriate well-known method. In order to approximate interval-valued fuzzy coefficients a hybrid scheme based on recurrent neural networks is introduced. A real data sets, on hydrology is used to test the applicability of the proposed method. The predictive ability of the model thus obtained is examined by cross-validation

کلیدواژه ها:

Fuzzy regression model ، Interval-valued fuzzy number ، Recurrent neural network ، Stability ، Convergence