Attention-Based Learning for Predicting Drug-Drug Interactions in Knowledge Graph Embedding Based on Multisource Fusion Information

Table 3

Ablation experimental results.

Method	KEGG-drug				OGB-biokg
	AUC	Acc	F1	AUPR	AUC	Acc	F1	AUPR
Ours	0.9705	0.9240	0.9257	0.9611	0.9959	0.9778	0.978	0.9953
W/o residual	0.9516	0.9005	0.9035	0.9292	0.9949	0.9729	0.9732	0.9943
W/o attention	0.9663	0.9161	0.9175	0.9555	0.9957	0.9765	0.9768	0.9950
W/o seq	0.9413	0.8894	0.8948	0.9146	0.9947	0.9717	0.9720	0.9939
W/o mole	0.9639	0.9125	0.9141	0.9514	0.9953	0.9749	0.9752	0.9947
W/o both	0.9442	0.8915	0.8921	0.9190	0.9952	0.9735	0.9722	0.9963