Neural network ensemble can significantly improve the generalization ability of learning systems through training a finite number of neural networks and then combining their results. It is not only helpful for experts to investigate machine learning and neural computing but also helpful for engineers to solve real world problems using neural network techniques. This paper studies the effectiveness of neural network ensemble in forecasting the nonlinear behavior of financial data and proposes a neural network model to forecast the foreign exchange rates. The experimental results show that the proposed model is able to achieve higher accuracy of the directional forecast. Besides the forecast model, a new model called AB neural network model is presented which is used to study the problems of configuration and training in neural networks. Finally, we apply the neural network into a real foreign exchange rate forecasting system. |