Improved sales time series predictions using deep neural networks with spatiotemporal dynamic pattern acquisition mechanism

作者：

Highlights：

• The proposed SDPA model incorporates 4 novel components for sales prediction.

• The SDK and HA components can better capture dynamic non-linear correlations from MTS.

• The SR component can detect linear correlations between time series dynamically.

• The DC component can provide more useful future information for the model training.

摘要

•The proposed SDPA model incorporates 4 novel components for sales prediction.•The SDK and HA components can better capture dynamic non-linear correlations from MTS.•The SR component can detect linear correlations between time series dynamically.•The DC component can provide more useful future information for the model training.

论文关键词：Sales prediction,Spatiotemporal dynamic kernel,Simultaneous regression,Hierarchical attention,Dynamic detection and alignment

论文评审过程：Received 8 January 2022, Revised 20 May 2022, Accepted 22 May 2022, Available online 14 June 2022, Version of Record 14 June 2022.

论文官网地址：https://doi.org/10.1016/j.ipm.2022.102987