MIANet: Multi-level temporal information aggregation in mixed-periodicity time series forecasting tasks

Sheng Wang, Xi Chen, Dongliang Ma, Chen Wang, Yong Wang, Honggang Qi, Gongjian Zhou, Qingli Li, Min Liu

Abstract: Highlights•MIANet thoroughly integrates multi-level temporal information in a compact manner.•We design a folded recurrent unit, circularly updating local and mini-local features.•We develop a novel decoder structure that strengthens the interplay of future vision.•We devise an SAR linear strategy that can practically solve the scale insensitivity.•Compared to eight other methods, MIANet achieves the best results on six datasets.