基于人工智能的人民币/美元汇率组合预测模型

Integrated forecasting model of RMB /USD exchange rate based on artificial intelligence

针对在人民币/美元汇率收益率的尖峰厚尾、波动性聚集性与杠杆效应等特征,单个预测模型往往难以完全将这些数据特点完全反应,为更有效地利用各个模型的优点,利用协整关系和神经网络的非线性特点将不同的单一模型进行组合。实证表明:组合模型能产生更好的预测精度。

In terms of such characteristics as high-spike with fat tail,volatility clustering and leverage effect of the RMB / USD exchange rate,a single prediction model is inadequate to complete the data characteristics. For more efficient use of the advantages of each model,the study integrates different single model by making use of cointegration relationship and single model nonlinear characteristic of neural network The results show that the integrated model is of higher prediction accuracy than the single model.