Table 4 Predicting the value of high-incidence days using XGBoost in the test data of the top-six population prefectures.
From: Machine learning algorithms for predicting days of high incidence for out-of-hospital cardiac arrest
AUROC (95%CI) | Accuracy | Sensitivity | Specificity | F1-score | |
|---|---|---|---|---|---|
Tokyo | 0.923 (0.905–0.938) | 0858 (0.835–0.879) | 0.841 (0.793–0.884) | 0.864 (0.838–0.888) | 0.732 (0.689–0.770) |
Kanagawa | 0.882 (0.860–0.904) | 0.804 (0.778–0.826) | 0.800 (0.746–0.844) | 0.805 (0.775–0.832) | 0.668 (0.625–0.707) |
Osaka | 0.888 (0.863–0.911) | 0.748 (0.721–0.774) | 0.890 (0.843–0.936) | 0.723 (0.694–0.751) | 0.514 (0.463–0.562) |
Aichi | 0.889 (0.863–0.912) | 0.783 (0.758–0.805) | 0.862 (0.818–0.904) | 0.761 (0.730–0.787) | 0.634 (0.589–0.674) |
Saitama | 0.879 (0.855–0.901) | 0.737 (0.710–0.762) | 0.879 (0.836–0.918) | 0.699 (0.667–0.728) | 0.585 (0.538–0.625) |
Chiba | 0.862 (0.831–0.891) | 0.761 (0.733–0.787) | 0.825 (0.774–0.874) | 0.745 (0.713–0.773) | 0.572 (0.520–0.617) |
