TABLE 16.

Combined features: performance and hyperparameters of graph models.

Model	Train_Acc	Val_Acc	Test_Acc	Train_F1	Val_F1	Test_F1	Hyperparameters
GraphSAGE_mean	95.136 ±0.07	90.183 ±0.069	86.656 ±0.209	0.9605 ±0.001	0.9532 ±0.003	0.8972 ±0.004	hidden dim=78, lr=0.0054, dropout=0.1975, Aggr=mean
GraphSAGE_max	93.04 ±0.65	88.57 ±0.655	84.593±0.499	0.9261 ±0.015	0.9111 ±0.025	0.9092±0.012	hidden dim=78, lr=0.0054, dropout=0.1975, Aggr=max
GATV2	93.30 ±0.28	89.68 ±0.88	83.78±1.013	0.9448 ±0.024	0.9359 ±0.014	0.8975±0.025	lr=0.012,heads=8, dropout=0.1, weight_decay=5e-4
GatConv	90.38 ±1.28	87.38 ±1.34	81.73±1.136	0.9524±0.012	0.948 ±0.013	0.8818±0.031	lr=0.073,heads=8, dropout=0.111, weight_decay=0.0005
Proposed Model	94.50 ±0.385	89.86 ±0.194	87.23 ±0.172	0.9642 ±0.001	0.9601 ±0.007	0.9509 ±0.004	lr=0.001, dropout=0.1 num_sage_layers=3, hidden_channel=33