Table 3 Experimental setup details.
Hardware/software | Details |
|---|---|
Platform | Google Colaboratory (high-RAM, GPU-enabled runtime) |
GPU | NVIDIA Tesla T4 (16GB VRAM) |
CPU | Intel Xeon (~ 2.2 GHz) |
RAM | 25.5 GB |
Python version | Python 3.10.12 |
Libraries and frameworks | - PyTorch 2.0- PyTorch Geometric 2.4.0- Scikit-learn- Pandas- NumPy |
Optimizer | Adam optimizer |
Learning rate (η) | 0.005 |
Weight decay | 10 − 410^{−4} |
Batch size | 512 |
Loss function | Weighted cross-entropy loss (w_churn = 10, w_(not churn) = 1) |
