Combining the advantages of fuzzy control and neural network, the thesis discusses how to an improved fuzzy Gaussian function neural network applied to numeric approximation of the searched system. The simulation results shows that this network can perform very well.