Estimation of fuzzy parameters based on neural networks using trapezoidal data

Fuzzy regression is a generalized regression model that shows the relationship between independent and dependent variables in the fuzzy environment. Fuzzy linear regression analysis is the generalization of regression models that is appropriate using all data based on a specific criterion. This paper uses an adaptive neural fuzzy inference system to analyze and predict a non-parametric fuzzy regression function with non-fuzzy inputs and symmetrical trapezoidal fuzzy outputs. To this end, a new hybrid algorithm is proposed in which fuzzy minimum squares and linear programming are used to optimize secondary weights. Algorithms are applied by multi layer validation to validate models. More precisely, the accuracy of the algorithms with simulations is fully confirmed. Finally, two simulation examples were used to examine the efficiency of the model, in which the data were defined as trapezoidal numbers and by teaching them and specifying the number of rules used, the unknown parameters were estimated.