CoDCL: Counterfactual Data Augmentation Contrastive Learning for Continuous-Time Dynamic Network Link Prediction

The rapid growth and continuous structural evolution of dynamic networks make effective predictions increasingly challenging. To enable prediction models to adapt to complex temporal environments, they need to be robust to emerging structural changes. We propose a dynamic network learning framework CoDCL, which combines counterfactual data augmentation with contrastive learning to address this deficiency.Furthermore, we devise a comprehensive strategy to generate high-quality counterfactual data, combining a dynamic treatments design with efficient structural neighborhood exploration to quantify the temporal changes in interaction patterns.Crucially, the entire CoDCL is designed as a plug-and-play universal module that can be seamlessly integrated into various existing temporal graph models without requiring architectural modifications.Extensive experiments on multiple real-world datasets demonstrate that CoDCL significantly gains state-of-the-art baseline models in the field of dynamic networks, confirming the critical role of integrating counterfactual data augmentation into dynamic representation learning.

💡 Research Summary

The paper introduces CoDCL (Counterfactual Data Augmentation Contrastive Learning), a novel plug‑and‑play framework designed to improve continuous‑time dynamic network link prediction. Existing temporal graph models largely rely on correlation‑based learning, which makes them vulnerable to structural changes, emerging nodes/edges, and distribution shifts between training and testing periods. CoDCL addresses these shortcomings by integrating causal reasoning with contrastive learning.

First, the authors define a treatment variable that captures the intensity of interaction between a node pair. Using a time‑constrained common‑neighbor count N(u,v,t) within a sliding window Δ, they compute a continuous interaction intensity ϕ(u,v) that can optionally incorporate exponential decay to emphasize recent events. A global percentile‑based threshold θ converts this continuous measure into a binary treatment T_uv(t) = I