Gate-based GWNet for process quality filter and multioutput prediction

Shifan Chen, Qunshan He, Peiyan Tu, Simengxu Qiao, He Zhang, Xinggao Liu

Abstract: Highlights•A spatiotemporal model for capturing complex fermentation patterns.•Graph Structure and Time Gating: Modules that capture trends and enable dynamic data filtering.•Dynamic Graph Updates: GCT and self-gated convolution for effective graph structure updates.•Gated Temporal Convolution: Self-gating in the output filtering for refined temporal feature adaptation.•Superior Performance: GWNet reduces MAE by 16.667% over MTGNN and Autoformer on three datasets.