Table 5 Ablation studies for our model
From: Medical code prediction via capsule networks and ICD knowledge
Methods | Macro F1 (%) | Micro F1 (%) | Test loss (%) | Top-10 recall (%) |
|---|---|---|---|---|
CapsNet | 21.8 | 64.7 | 3.5 | 76.1 |
BiCapsNet | 23.1 | 67.0 | 3.1 | 81.9 |
BiCapsNetLE | 26.5 | 67.5 | 2.9 | 82.3 |