The application of gradient evolution algorithm to an intuitionistic fuzzy neural network for forecasting medical cost of acute hepatitis treatment in Taiwan.

Forecasting is important in the decision-making process. Therefore, among many forecasting algorithms that have been developed, fuzzy neural network (FNN) has been widely applied. The FNN structure is commonly trained using a back-propagation algorithm. However, this algorithm is sensitive to the initial weights and high computation, especially for complex problems. Thus, this study intended to overcome these drawbacks by applying the gradient evolution (GE) algorithm to train the intuitionistic FNN (IFNN). The proposed algorithm, GEIFNN, was verified using ten benchmark datasets. The results were compared with some other metaheuristic-based IFNN algorithms, such as genetic algorithm, particle swarm optimization algorithm, and differential evolution algorithm. The computational results show that the proposed GEIFNN relatively outperformed other tested algorithms, especially in training error. Furthermore, the proposed algorithm was applied to forecast medical costs for the treatment of acute hepatitis in Taiwan. The result also shows that GEIFNN can obtain the smallest training error. • Improve backpropagation on intuitionistic fuzzy neural network using a gradient evolution algorithm. • Verify the proposed algorithm with ten benchmark datasets. • Computational results show that the proposed algorithm gives promising results. • Apply the proposed algorithm to forecast medical cost for acute hepatitis treatment in Taiwan. • The contributions are developing GEIFNN algorithm and solving a real problem. [ABSTRACT FROM AUTHOR]