Adaptive imbalanced node classification graph contrastive learning

Graph Contrastive Learning (GCL) is a powerful self-supervised technique for learning node and graph representations. However, real-world graph data often exhibit imbalanced class distributions, which pose significant challenges to GCL’s effectiveness. Our experiments show that current state-of-the-art (SOTA) methods perform poorly under imbalanced settings. To address this, we propose a novel GCL framework called AIGCL for imbalanced node classification. This framework automatically and adaptively balances the node representations learned by GCL. Specifically, we introduce a new data augmentation method that retains more information from minority class nodes during graph augmentation. Additionally, we use an imbalance rate adaptive sampling strategy to balance the data. We also incorporate a Variational Graph Autoencoder (VGAE) with an encoder–decoder structure to pretrain the data and generate high-quality pseudo-labels. Our experiments demonstrate that under imbalanced settings, our model improves classification accuracy by 4 %-12 % compared to baseline models, significantly enhancing the performance of minority class nodes. This figure shows a Graph Contrastive Learning process, including three main modules. The GCL Model module demonstrates our graph augmentation method for retaining more minority class information and uses a graph encoder and loss function to learn graph representations. The Pre-training module uses the trained Encoder–Decoder model to generate high-quality embeddings and generate pseudo-labels through K-means clustering. The Resampling module balances the node class distribution through mixed sampling techniques, including oversampling and undersampling. • We introduce a VGAE with an encoder-decoder structure to learn graph structure and node features, improving robustness and generalization for pseudo-label generation. • We design a graph-tailored resampling strategy. We prioritize removing low-importance nodes during undersampling and synthesizing high-importance nodes during oversampling. • We create a multi-level, step-by-step sampling strategy. A hybrid sampling method adjusts class proportions, and an augmentation function retains minority node information. • We develop a new data augmentation technique that prioritizes keeping minority class node information while masking majority class nodes, helping the model capture minority features.