摘要
本文选取影响人民币汇率波动的有关结构变量,分别通过线性MA模型和基于遗传算法改进的GABP神经网络模型,对人民币汇率波动进行模拟和预测。通过比较发现,汇率缺乏弹性时期,逐月MA模型的历史拟合和样本外预测效果最优;随着汇率改革的不断推进和汇率弹性化的增强,GABP神经网络模型在汇率波动的模拟和预测方面均有最优表现,故汇率波动预测模型应随汇率弹性及其波动特性不同因时制宜。同时结果表明,汇率弹性化能够加深汇率波动及其结构变量间的均衡关系,利率市场化改革应与汇率市场化改革协调推进。
This article selects the structural variables related with RMB exchange rate volatility to simulate and forecast the exchange rate volatility through linear MA, GABP neural network model improved on genetic algorithm respectively. By comparison, when there is a lack of exchange rate elasticity, the history matching and sample fore-casting performance of monthly MA model are optimal. Along with the promotion of exchange rate reform and increase in its elasticity, GABP network model has done a good job both in simulation of exchange rate volatility and its fore- cast. Therefore, different forecasting models of exchange rate volatility should be adopted according to the features of the elasticity in different periods. Meanwhile, the model results show that the exchange rate elastification can deepen the equilibrium between exchange rate volatility and its structural variables. And the interest rate liberalization reform should coordinate with the liberalization of exchange rate reform.
出处
《金融发展研究》
2013年第4期8-13,共6页
Journal Of Financial Development Research
关键词
汇率弹性化
汇率波动
神经网络
exchange rate elastification, exchange rate volatility, neural network
