信号分解与融合神经网络的金融数据预测研究  

Research on Financial Data Forecast Based on Signal Decomposition and Fusion Neural Network

作　　者：金丰 邵清[1] JIN Feng;SHAO Qing(School of Optoelectronic Information and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《小型微型计算机系统》2020年第6期1140-1146,共7页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61703278)资助;国家重点研发计划项目(2018YFB1700902,2018YFB1702601)资助.

摘　　要：针对目前特征+模型的预测方案在金融时序数据上预测精度不足的问题,提出一种基于变分模态分解的长短期记忆神经网络-因子分解机双通道融合模型.首先利用变分模态分解分解金融时序数据,得到能表现时序数据趋势和变化信息的模态,接着将模态信息输入到长短期记忆神经网络-因子分解机双通道融合模型中,长短期记忆神经网络得到数据的时序特征,因子分解机得到数据的交互特征,通过改进的局部连接层融合,使模型能同时表达数据的时序特征和交互特征,提升时序数据预测的效果.以沪深300指数和标普500指数的数据为例,对所提模型进行验证,结果表明其在预测性能和泛化性能上都有明显的优势.Aiming at the problem of insufficient prediction accuracy of the current feature+model prediction scheme on financial time series data,a long short-term memory neural network-Factorization machine dual channel fusion model based on variational modal decomposition.First,use the variational modal decomposition(VMD)to decompose financial time series data to obtain a modal that can represent the trend and change information of the time series data,and then input the modal information to the long short-term memory neural network-Factorization machine dual channel fusion model,long short-term memory neural network obtains the time series characteristics of data,and Factorization machine obtains the interactive characteristics of data,combine them together through improved Locally-Connected Layer,the model can simultaneously express the time series characteristics and interaction characteristics of the data,improving the effect of time series data prediction.Taking the data of Shanghai and Shenzhen 300 Index and S&P 500 Index as examples,the proposed model is verified,and the results show that it has obvious advantages in predictive performance and generalization performance.

关 键 词：金融时序数据预测 变分模态分解 长短期记忆神经网络 因子分解机 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]