摘要
本文将人工神经网络应用于汇率预报.应用从1987年5月至1992年12月伦敦和纽约两大外汇市场马克对美元的市场即期汇率数据,建立前向组合神经网络预报模型.训练后的神经网络不仅能准确地拟会汇率的过去值,而且能较精确地预报汇率的未来趋势.计算结果表明:汇率的神经网络预报方法比统计预报方法优越.
This paper presents a neural networks approach to exchange rates analysis. Real observations of exchange rate (DM/$) in two exchange markets has been as a benchmark in our experiments. Feedforward combined networks have been designed to model exchange rates over the period from May 1987 to December 1992 weekly for the foreign exchange markets of London and New York. Remarkable success has been achieved in training the networks to learn the exchange rate curve for each of these markets and in making accurate predictions. Our results show that the neural network approach is a leading contender with the statistical modelling approachs.
出处
《系统工程理论与实践》
EI
CSCD
北大核心
1996年第6期13-20,29,共9页
Systems Engineering-Theory & Practice
关键词
汇率
汇率预报
人工神经网络
多步预报
exchange rate
neural networks
conjugate gradient
time series models
training
one-lag prediction
multi-lag prediction
combined modeling
forecasting