首页> 外国专利> Ensemble of grouped two-stage attention-based recurrent neural networks for multivariate time series prediction

Ensemble of grouped two-stage attention-based recurrent neural networks for multivariate time series prediction

机译:基于分组两阶段注意的多变量时间序列预测递归神经网络集成

摘要

A method for multivariate time series prediction is created. Each time series consisting of a set of multiple drive time series and a target time series is broken down into a raw component, a mold component and a trend component (320). For each disassembled component, a control time series relevant to it isselected from the set and hidden features of the selected drive time series are obtained by applying (330) the set to an attention-based input encoder of an ensemble of grouped two-stage attention-based recurrent neural networks (EC-DARNNS). The hidden features are displayed in a hidden space using an attention-based temporal decoEC-DARNNS groups automatically (330). Each clustered two-stage attention-based RNN in the ensemble is firmly assigned to a respective disassembled component and is applied to it. A particular value of one or more future time steps for the target series is predicted by the EC-DARNNS on the basis of the respective forecast outputs for each disassembled component (330).
机译:提出了一种多元时间序列预测方法。由一组多驱动时间序列和一个目标时间序列组成的每个时间序列被分解为原始组件、模具组件和趋势组件(320)。对于每个分解的部件,从集合中选择与其相关的控制时间序列,并通过将(330)集合应用于分组的基于两阶段注意的递归神经网络(EC-DARNS)集合的基于注意的输入编码器,获得所选驱动时间序列的隐藏特征。隐藏特征使用基于注意的时间解码自动显示在隐藏空间中(330)。集合中的每一个基于两阶段注意的集群RNN都被牢固地分配给各自的分解组件,并应用于该组件。目标序列的一个或多个未来时间步长的特定值由EC-DARNS基于每个分解部件(330)的各自预测输出进行预测。

著录项

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号