A BP Neural Network Activation Function Used in Exchange Rate Forecasting

Traditional BP algorithm has defects of local minima and slow convergence. Based on application for exchange rate forecasting, this paper gives an improved BP neural network activation function. Then propose a new activation function after by analyze influence factor of the exchange rate and design structure of forecasting of the improved BP model. Take simulation forecast for a sample data of the RMB exchange rate. The results show that the improved BP neural network not only has accelerated in the training speed, also has a significant improvement in forecasting performance compare to the traditional BP neural network.