A modular design of fuzzy neural network with application to telecommunication network management control

There were several attempts recently in making a combination of fuzzy logic and neural networks for better performance in expert systems. Since all operations involve fuzzy set theory, the amount of calculation per inference increases dramatically. In this paper, a modular fuzzy neural network model for expert systems, based on a multilayer perceptron, is proposed. The model is capable of handling both the crisp and the fuzzy inputs, and the output of which is the inferred satisfactory solution. The effectiveness of the model is tested on the telecommunication network management control problem. The model is suitable for designing classification-type expert systems in which the domain knowledge may be a poor combination of knowledge from different experts.