Abstract:
Intuitionistic fuzzy inference systems (IFISs) incorporate imprecision in the construction of membership functions present in fuzzy inference systems. In this paper we design intuitionistic neuro-fuzzy networks to adapt the antecedent and consequent parameters of IFISs. We also propose a mean of maximum defuzzification method for a class of Takagi-Sugeno IFISs and this method is compared with the basic defuzzification distribution operator. On both real-life credit scoring data and seven benchmark regression datasets we show that the intuitionistic neuro-fuzzy network trained with particle swarm optimization outperforms traditional ANFIS methods (hybrid and backpropagation) and ANFIS trained with evolutionary algorithms (genetic algorithm and particle swarm optimization), respectively. A set of nonparametric tests for multiple datasets is performed to demonstrate statistical differences between the algorithms. In the task of adapting the intuitionistic neuro-fuzzy network, we show that particle swarm optimization provides a higher prediction accuracy compared with traditional algorithms based on gradient descent or least-squares estimation.