Multi-step forecasting of multivariate time series using multi-attention collaborative network

作者：

Highlights：

• A triangle structured multi-attention model for multivariate time series prediction.

• Variables-distillation attention selects the most relevant non-predictive variables.

• The KeLSTM uses non-predictive variables to enrich the target variable’s semantics.

• The model distinguishes the effects of target and non-predictive variables on tasks.

• MACN is competitive in multivariate time series multi-step prediction.

摘要

•A triangle structured multi-attention model for multivariate time series prediction.•Variables-distillation attention selects the most relevant non-predictive variables.•The KeLSTM uses non-predictive variables to enrich the target variable’s semantics.•The model distinguishes the effects of target and non-predictive variables on tasks.•MACN is competitive in multivariate time series multi-step prediction.

论文关键词：Multivariate time series,Multi-attention,Deep neural network,Multi-step forecasting

论文评审过程：Received 29 January 2022, Revised 13 July 2022, Accepted 9 August 2022, Available online 17 August 2022, Version of Record 26 August 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.118516