CRRL: Contrastive Region Relevance Learning Framework for Cross-city Traffic Prediction

Genan Dai, Wenfeng Yi, Jinzhou Cao, Zhaoya Gong, Xianghua Fu, Bowen Zhang

Abstract: Highlights•CRRL framework enables cross-city traffic prediction under data scarcity.•Dual-branch encoder captures local and high-order spatiotemporal relations.•Pseudo-label Generation aligns cross-city patterns via embedding similarities.•Reliable region selection minimizes negative transfer with contrastive learning.•State-of-the-art performance through robust region-level knowledge transfer.