This is the Graph Transformer solution for DGraphFin leaderboard, which is the Top-2 solution until July, 2023.
python = 3.8
torch = 1.10.1+cu102
torch-geometric = 2.0.4
scikit-learn = 0.20.4
torch-scatter = 2.0.9
torch-sparse = 0.6.12
The dataset DGraph-Fin should be download and placed under the folder ./dataset/DGraph/raw
.
python main.py --model transformer --epochs 1000 --runs 10
Performance on DGraph-Fin (10 runs):
Best Valid AUC: 0.7717 ± 0.0017
Best Test AUC: 0.7838 ± 0.0015