Table 1 Testing-set predictive performance of MNL and XGBoost.

From: Modeling protective action decision-making in earthquakes by using explainable machine learning and video data

Protective action

Share of obs.

Precision

Recall

F1-score

MNL

XGBoost

MNL

XGBoost

MNL

XGBoost

Other

0.556

0.717

0.978

0.798

1.000

0.755

0.989

Drop and Cover

0.269

0.587

0.909

0.628

0.930

0.607

0.920

Hold on

0.088

0.000

0.833

0.000

0.714

0.000

0.769

Evacuate

0.087

0.600

1.000

0.643

0.929

0.621

0.963

  1. Overall Accuracy: MNL: 0.669, XGBoost: 0.950.