Table 1 Performance of pediatric emergency prediction using natural language processing techniques and topic models.

From: A pediatric emergency prediction model using natural language process in the pediatric emergency department

Variable

Logistic regression

XGBoost

Gradient boosting

Random forest

KM-BERT

KM-BERT with MLM*

AUROC

0.698 ± 0.002

0.714 ± 0.002

0.715 ± 0.002

0.680 ± 0.002

0.788 ± 0.002

0.839 ± 0.001

AUPRC

0.752 ± 0.002

0.776 ± 0.001

0.778 ± 0.001

0.735 ± 0.002

0.837 ± 0.002

0.879 ± 0.001

Recall

0.629 ± 0.003

0.618 ± 0.002

0.626 ± 0.003

0.625 ± 0.002

0.719 ± 0.002

0.724 ± 0.002

Precision

0.728 ± 0.002

0.741 ± 0.001

0.737 ± 0.001

0.716 ± 0.002

0.775 ± 0.002

0.829 ± 0.001

F1-score

0.675 ± 0.002

0.674 ± 0.002

0.677 ± 0.001

0.667 ± 0.001

0.746 ± 0.002

0.773 ± 0.001

Accuracy

0.643 ± 0.002

0.648 ± 0.002

0.649 ± 0.001

0.633 ± 0.002

0.712 ± 0.002

0.749 ± 0.001

Brier

0.215 ± 0.000

0.209 ± 0.000

0.209 ± 0.000

0.225 ± 0.001

0.188 ± 0.001

0.164 ± 0.001

  1. KM-BERT, Korean medical bidirectional encoder representations from transformers; MLM, masked language modeling; AUROC, area under the receiver operating characteristics; AUPRC, area under the precision-recall curve.
  2. Bold values indicate the best-performing model across all metrics.; *Indicates overall best-performing model.
Back to article page