基于双神经网络通道的时间序列预测框架  

Time Series Prediction Framework Based on Dual Neural Network Channel

作　　者：吴双双 宋恺涛 陆建峰[1] WU Shuangshuang;SONG Kaitao;LU Jianfeng(School of Computer Science and Engineering,Nanjing University of Science and Technology,Nanjing 210094)

机构地区：[1]南京理工大学计算机科学与工程学院,南京210094

出　　处：《舰船电子工程》2018年第9期84-89,共6页Ship Electronic Engineering

基　　金：江苏省自然基金项目"基于多模态数据的脑功能网络分析技术研究"(编号:BK20131351)资助

摘　　要：为了提高时间序列预测方法的预测精度,提出了一种基于双神经网络通道的时间序列预测框架。双通道框架组合了卷积神经网络和循环神经网络,卷积神经网络通道利用卷积和池化操作提取出时间序列的深层特征,循环神经网络通道能够提取出长序列依赖特征。此外,针对循环神经网络通道,设计了基于注意力机制的改进,提高预测精度。在EEM2016能源价格预测比赛提供的数据集上进行了实验,实验结果表明所提双通道框架可以取得比单通道框架更高的预测精度。In order to improve the prediction accuracy of the time series prediction method,a time series prediction framework based on the dual neural network channel is proposed.The dual-channel framework combines convolutional neural network and re-current neural network.The convolutional neural network channel can extract deep features of time series by convolution and pooling operation,and the recurrent neural network channel can extract long sequence dependent features.In addition,in view of the recur-rent neural network channel,an improvement based on attention mechanism is designed to improve the prediction accuracy.Experi-ments on the data set provided by EEM2016 energy price forecast competition are made,and the results show that the proposed du-al-channel framework can obtain higher prediction accuracy than single channel framework.

关 键 词：卷积神经网络 循环神经网络 双通道框架 时间序列预测 注意力机制 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]