基于VMD-EEMD-LSTM模型的沪深300指数预测研究

Research on the Forecast of CSI 300 Index Based on VMD-EEMD-LSTM Model

股指序列的非平稳、非线性和长记忆等特性使其预测难度较大,为了改善已有模型的预测精度,本文首次提出一种新的融合二层分解技术及长短期记忆深度神经网络的沪深300股指收益率组合预测模型,该组合模型由变分模态分解(VMD)、集合经验模态分解(EEMD)两种分解技术及长短期记忆神经网络(LSTM)组成。选取沪深300指数2016年3月1日至2019年8月30日,共858个交易数据作为建模对象,并采用后50个交易日数据作为测试样本。实证结果表明,本文提出的VMD-EEMD-LSTM组合模型预测评估指标RMSE、MAE、MAPE取值分别为0.5532、0.4613、1.2842,显著优于已有二次分解及单次分解组合模型,具有显著的预测优势。本文的研究结果可助益于金融市场监管当局及时开展风险预警工作,制定合理的风险管理政策;同时也可帮助投资者采取有效的投资策略,降低投资风险。

The non-stationary,non-linear,long memory and other characteristics of the stock index series make it difficult to predict.In order to improve the prediction accuracy of existing models,this paper combines a two-layer decomposition technique and a long-term and short-term memory deep neural network,and proposes a new forecasting model of Shanghai and Shenzhen 300 stock index returns for the first time.The combined model is composed of Variational Modal Decomposition(VMD),Ensemble Empirical Mode Decomposition(EEMD)and two long-term and short-term memory neural networks(LSTM).A total of 858 transaction data from the CSI 300 Index from March 1,2016 to August 30,2019 were selected as the modeling objects,and the data from the last50 trading days were taken as the test sample.The empirical results show that the RMSE,MAE and MAPE of the VMD-EEMD-LSTM model are 0.5532,0.4613,and 1.2842,respectively,which are significantly better than the existing secondary decomposition and single decomposition combination models,and have significant prediction advantages.The research results can help the financial market regulatory authorities to carry out risk early warning work in time and formulate reasonable risk management policies,at the same time,it can also help investors adopt effective investment strategies and reduce investment risks.