<p>With the increasing complexity and temporal nature of real-world data, temporal knowledge graph completion has emerged as a research hotspot. This paper introduces a novel approach named temporal counterfactual augmentation (TCA) to address the challenges of temporal knowledge graph (TKG) completion, particularly the issues of temporal dynamics and data scarcity. Specifically, our TCA method leverages counterfactual augmentation to model the temporal community structure and its impact on entity relationships, thereby significantly enhancing the model’s ability to predict missing temporal links. By constructing “what-if” scenarios, our method provides a deeper understanding of TKGs, resulting in improved predictive performance and a more nuanced representation of temporal changes. Extensive experimental evaluations validate the effectiveness of our TCA approach, demonstrating its robustness to data sparsity and noise, as well as its clear superiority in TKG completion tasks.</p>