Abstract: Using the method of Radial- Basis Function artificial neural networks and MATLAB toolbox combined with the meteorological data of the average temperature, the sunshine hours, the rainfall amount, the article has established the RBF neural network system to forecast the degree of pest occurrence in Xinjiang Shihezi areas. We has confirmed the forecast accuracy of the system through the examples, and made comparison between the RBF network and the commonly used BP network in which tile training sample simulation of the two has revealed the RBF network is obviously more precise than the BP network. To time the procedure shows the training time of RBF network is 1.2030s which is much shorter than BP training needed,Therefore the RBF neural network has a very promising application value.