Bahareh Arefmanesh - Islamic Azad University- Science and Research Unit
Teshnehlab - Departemant of Electric Engineering, Khajeh Nasir Toosi of Technology Tehran, Iran
Shenasa - Departemant of Electric Engineering, Khajeh Nasir Toosi of Technology Tehran, Iran
Kamali - Iran Meteorology Organization

Combination of neural networks and fhzzy systems yields intelligent Fuzzy-Neural system used to solve problems such as modeling, prediction, etc. In this paper, Fuzzy- Neural Network (FNN) has been designed to predict sea waves that are non-linear phenomena. The system trains parameters in three phases: first, system trains conclusion parameters (center of output membership fi~nction);t hen, the system trains antecedent parameters (center of input membership function) with conclusion parameters and; finally, the system trains variance of membership function with other parameters. Result of each phase was compared with each other. Result of third phase is better than other phases. The performance of this phase is slower than other phases because it has higher degrees of freedom. Next stage, three previous phases would be used recurrent Fuzzy- Neural Network to train the parameters. Training conclusion parameters, antecedent parameters and variance parameters together has the best result. In this paper, Error Back Propagation Algorithm was used for learning mechanism, and Fuzzy-Neural Networks were evaluated by Mean Square of Error (MSE).

Fuzzy-neural Networks, Error Back Propagation Algorithm, and Recurrent Fuzzy-neural Networks