Nonlinear Combination Forecasting Model and Application Based on Radial Basis Function Neural Networks

According to the sales forecasting problems of financial equipments,on the basis of radial basis function (RBF, for short) theory, a nonlinear combination forecasting model is established based on RBF neural network by transforming a nonlinear mapping(input layer to hidden layer) to linear mapping in another space, this model greatly improves learning speed of neural networks, avoids local optimum and overcomes the disadvantages of low accuracy for linear combination forecasting in some forecasting points. On the basis of the industrial characteristics of the products, the forecasting results of traditional seasonal index smoothness and BP neural network forecasting model are taken as an input vector of RBF neural network, and the actual value of the respective moment is taken as an output. Network is trained by enough forecasting samples to achieve a high accuracy. Finally, the sales forecasting of JL106 money binding machine is taken as an example to illustrate the effectiveness of the model.