Table 4.

Performance comparison with the state-of-the-art methods on three tasks of Dataset2

	ACC	AUPR	AUC	F1	Precision	Recall
Task1
MDDI-SCL	0.9516	0.9862	0.9995	0.9321	0.9162	0.9500
MDF-SA-DDI	0.9291	0.9773	0.9996	0.9117	0.9381	0.8910
DDIMDL	0.9229	0.9637	0.9993	0.9105	0.9212	0.9039
Lee et al.'s methods	0.9370	0.9791	0.9991	0.9181	0.9226	0.9153
DeepDDI	0.7211	0.7724	0.9914	0.6854	0.6654	0.7183
DNN	0.7908	0.8539	0.9949	0.7649	0.7560	0.8046
RF	0.6956	0.7567	0.9892	0.5760	0.6694	0.5426
KNN	0.5797	0.5964	0.8998	0.3805	0.4758	0.3347
LR	0.5229	0.5288	0.9805	0.2373	0.3128	0.2185
Task2
MDDI-SCL	0.6595	0.6794	0.9757	0.5578	0.5605	0.5712
MDF-SA-DDI	0.6664	0.6820	0.9862	0.5919	0.6526	0.5518
DDIMDL	0.6720	0.7086	0.9885	0.5817	0.6680	0.5295
Lee et al.'s methods	0.6917	0.7119	0.9687	0.5934	0.6144	0.5848
DeepDDI	0.5883	0.5851	0.9746	0.4709	0.5250	0.4361
DNN	0.6687	0.6838	0.9818	0.6164	0.7279	0.5479
Task3
MDDI-SCL	0.4696	0.4261	0.9315	0.2838	0.3160	0.2773
MDF-SA-DDI	0.4794	0.4450	0.9686	0.2937	0.3667	0.2659
DDIMDL	0.4699	0.4386	0.9685	0.3032	0.3773	0.2729
Lee et al.'s methods	0.4867	0.4349	0.9093	0.3082	0.3355	0.3066
DeepDDI	0.3611	0.2820	0.9264	0.1868	0.2301	0.1711
DNN	0.4570	0.4129	0.9565	0.2997	0.4345	0.2508