Signed graph embedding via multi-order neighborhood feature fusion and contrastive learning

Chaobo He, Hao Cheng, Jiaqi Yang, Yong Tang, Quanlong Guan

Abstract: Highlights•A multi-order signed graph convolution network (MOSGCN) is proposed.•MOSGCN is powered by the structural balance theory.•MOSGCN can adaptively fuse neighborhood features from different orders.•MOSGCN is trained via special signed graph contrastive learning framework.•MOSGCN outperforms state-of-the-art baselines.