基于CEEMDAN-GAN的金融时间序列预测建模研究

Research on Financial Time Series Forecast Modeling Based on CEEMDAN-GAN

近年来,深度学习模型不断地被应用于时间序列的预测之中,而且效果显著。文章将自适应噪声完备集合经验模态分解(CEEMDAN)和随机森林算法引入到时间序列预测建模之中,提出了一种基于CEEMDAN-GAN的金融时间序列预测模型;以美元兑人民币汇率为测试数据,运用CEEMDAN对原始数据进行分解和重要性评估,通过不断优化内涵模态分量(IMF)的重构方式提高模型预测准确率;将CEEMDAN-GAN与当下流行的时间序列预测模型LSTM、GRU相比较。实验结果表明,CEEMDAN-GAN的预测结果优于现有的时间序列预测建模方法,具有更好的解释能力和更低的预测误差。

In recent years,deep learning models have been continuously applied to time series prediction,and the effect is remarkable.The adaptive noise complete ensemble empirical mode decomposition(CEEMDAN) and the random forest algorithm are introduced into the time series prediction modeling,and a financial time series forecasting model based on CEEMDAN-GAN is proposed.Taking the exchange rate of USD and RMB as the test data,CEEMDAN is used to decompose and evaluate the importance of the original data,and the prediction accuracy of the model is improved by continuously optimizing the reconstruction method of the Intrinsic Modal Component(IMF).Compared with the time series forecasting models LSTM and GRU,the experiments show that the forecasting results of CEEMDAN-GAN are better than the existing time series forecasting modeling methods,and have better explanatory ability and lower forecasting errors.