Enhancing few-sample spatio-temporal prediction via relational fusion-based hypergraph neural network

Xiaocao Ouyang, Yanhua Li, Dongyu Guo, Wei Huang, Xin Yang, Yan Yang, Junbo Zhang, Tianrui Li

Abstract: Highlights•A relational fusion-based hypergraph network is proposed for few-sample prediction.•A learnable matrix is trained to model dynamic high-order spatial relations.•A mixed-order relation-aware network is designed to learn spatial relationships.•A hierarchical module explores micro- and macro-level temporal relations.•Our proposed model achieves superior results in traffic flow, speed, and air quality.