Table 2:

Results for the MIMIC II dataset.

			F1		AUC (PR)		AUC (ROC)		P@k		R@k
	Prec.	Recall	Micro	Macro	Micro	Macro	Micro	Macro	8	40	8	40
Flat SVM (Perotte et al., 2013)	0.867	0.164	0.276	–	–	–	–	–	–	–	–	–
Hier. SVM (Perotte et al., 2013)	0.577	0.301	0.395	–	–	–	–	–	–	–	–	–
Logistic (Vani et al., 2017)	0.774	0.395	0.523	0.042	0.541	0.125	0.919	0.704	0.913	0.572	0.169	0.528
Attn BoW (Vani et al., 2017)	0.745	0.399	0.52	0.027	0.521	0.079	0.975	0.807	0.912	0.549	0.169	0.508
GRU–128 (Vani et al., 2017)	0.725	0.396	0.512	0.027	0.523	0.082	0.976	0.827	0.906	0.541	0.168	0.501
BiGRU–64 (Vani et al., 2017)	0.715	0.367	0.485	0.021	0.493	0.071	0.973	0.811	0.892	0.522	0.165	0.483
GRNN–128 (Vani et al., 2017)	0.753	0.472	0.58	0.052	0.587	0.126	0.976	0.815	0.93	0.592	0.172	0.548
BiGRNN–64 (Vani et al., 2017)	0.761	0.466	0.578	0.054	0.589	0.131	0.975	0.798	0.925	0.596	0.172	0.552
CNN (Baumel et al., 2018)*	0.810	0.403	0.538	0.031	0.599	0.127	0.971	0.759	0.931	0.585	0.207	0.586
Matching Network*	0.439	0.388	0.412	0.014	0.394	0.034	0.893	0.551	0.793	0.427	0.172	0.425
Match–CNN (Ours)	0.605	0.561	0.582	0.064	0.612	0.148	0.977	0.792	0.930	0.586	0.207	0.590
Match–CNN Ens. (Ours)	0.616	0.567	0.591	0.066	0.623	0.157	0.977	0.793	0.935	0.594	0.208	0.598

Models marked with * represent our custom implementations.