摘要
神经网络在理论上具有无限的函数逼近能力,它在预测领域可以取得很好的效果。利用神经网络的数值逼近与记忆功能,根据汇率历史观测数值,可以识别出汇率序列的内在模式。本文首先说明了利用神经网络进行汇率预测的原理和方法,然后着重探讨了神经网络汇率预测的重要步骤,最后根据不同的衡量指标对测试结果进行了分析。
Neural network is able to approach any functions infinitely in theory. And there is a good effect in forecasting field with neural network. With the ability of numeric approaching and memory, it can identify the inherence pattern of exchange rate sequence according to the historic data. At the beginning. This paper first introduces the basic principles of exchange rate forecasting with neural network. Secondly, the steps of forecasting are being discussed in detail. And at last, a software is been implemented as a forecaster of exchange rate and there are some analyses of the forecasted results using different evaluating methods with the software.
出处
《计算机与现代化》
2006年第2期105-108,共4页
Computer and Modernization
关键词
神经网络
汇率
预测
BP
neural network
exchange rate
forecast
back propagation
