Skip to main content

Table 3 Results on MIMIC-III, 50 labels

From: An explainable CNN approach for medical codes prediction from clinical text

Model

AUC

F1

P@5

Macro

Micro

Macro

Micro

C-MemNN [11]

0.833

0.42

Shi et al. [13]

-

0.900

-

0.532

-

CAML [12]

0.875

0.909

0.532

0.614

0.609

Logistic regression

0.828

0.862

0.477

0.530

0.545

SWAM-CAML

0.900*

0.924*

0.593

0.648

0.625*

SWAM-textCNN

0.892

0.919

0.603*

0.652*

0.620

  1. (*) by the bold (best) result indicates significantly improved results compared to the other methods, the bootstrapping method [30] is used for the statistical significance analysis, \(p<0.01\)