Simultaneous Optimization of Effective Factors on Artificial Neural Network Performance Using Box-Behnken Design and Fuzzy Programming

Tuning parameters has the most important effect in performance of Neural Network. Main problem in using of ANN is parameter tuning because there is no definite and explicit method to select optimal parameters for the ANN parameters. In this study, three artificial neural network performance criteria and also three important factors which affect the selected criteria have been introduced. Moreover, Box-Behnken design has been applied to analyze the ANN structure parameters and its performance. The proposed approach has been implemented for a simulated process according to a complex mathematical function. After extracting relation between controllable factors and performance criteria, fuzzy programming is used for finding the optimal combination of controllable factors in order the best performance of Neural Network. Results attained from the numerical example show efficiency of the proposed method. Generally, proposed approach can be used for tuning of neural network’s parameters in other problems.