Similar to node classification, edge classification is also a crucial research topic in graph learning. While existing studies predominantly focus on imbalanced node classification, research on imbalanced edge classification remains relatively scarce. Nevertheless, in practical applications, there are numerous tasks that require classifying different types of relationships, such as identifying illegal transactions. In such scenarios, the number of illegal transaction samples is often significantly smaller than that of normal transactions. Directly applying GNN classifiers in these cases can lead to inadequate feature learning for minority-class samples, thereby reducing overall model performance. Thus, there is a need to design algorithms specifically for imbalanced edge classification in GNN models. Drawing inspiration from research on imbalanced node classification, we propose the EdgeSMOTE algorithm. We evaluate our algorithm on two benchmark datasets, and the results demonstrate its superior performance compared to baseline methods.