基于鲁棒加权非线性组合模型的人民币汇率预测研究
RMB Exchange Rate Forecasting Based on a Hybrid Nonlinear Model with Robust Weighted Mechanism
综合考虑相关模型的特征和人民币汇率的非线性性质,构建一个基于ARIMA模型、SVM模型和BP神经网络模型的鲁棒加权非线性组合模型,通过该模型对人民币汇率进行预测。首先利用ARIMA模型从线性特征方向对人民币汇率时间序列进行拟合及预测,然后分别通过SVM模型和BP神经网络模型从非线性特征方面对人民币汇率时间序列进行拟合及预测,最后利用鲁棒加权机制将各模型的预测结果进行非线性组合,从而得到最终预测值。该方法不仅综合考虑了人民币汇率的内在特征和单个模型的不同特性,而且考虑了单个模型之间的关联性。实证结果表明:非线性组合模型较单个模型及线性组合模型预测效果更理想,更适合于类似人民币汇率等具有特殊性质的时间序列的预测。
ombined with the ARIMA model, the SVM model and the characteristic of the BP neural network model and the nonlinear properties of the RMB exchange rate, this paper develops a nonlinear combination model with a robust weighted mechanism. For the linearity of RMB exchange rate, the time series was fitted and forecasted with ARIMA model; For the nonlinearity of RMB exchange rate, the time series was fitted and forecasted with SVM model and BPNN model respectively. The prediction result of the original time series was combined nonliearily by robust weighted mechanism. This method not only considers the intrinsic characteristics of the RMB exchange rate and the different characteristics of a single model, but also considers the correlation between single models. The empirical results show that the prediction performance of nonlinear combined model is better than that of the single model and linear combination model, and is suitable for prediction of special time series, such as the RMB exchange rate.
毛舟、谢赤
财政、金融
人民币汇率RIMA模型SVM模型BP神经网络模型非线性组合模型
RMB exchange rateARIMA modelSVM modelBPNN modelnonlinear combined model
毛舟,谢赤.基于鲁棒加权非线性组合模型的人民币汇率预测研究[EB/OL].(2015-05-19)[2025-08-18].http://www.paper.edu.cn/releasepaper/content/201505-246.点此复制
评论