Table 2 F1-score comparison using InSAR data across earthquake events

From: Scalable variational learning for noisy-OR Bayesian networks with normalizing flows for complex cascading disaster systems

Model

Landslides

Liquefaction

Building Damage

 

HT

PR

HK

HT

PR

HK

HT

PR

HK

DisasterVINF

0.9281

0.9213

0.9192

NGA

0.9154

NGA

0.9285

0.9222

NGA

Prior Model

0.8813

0.8624

0.8410

NGA

0.8218

NGA

0.7853

0.6722

NGA

VBCI

0.8929

0.8756

0.8825

NGA

0.8857

NGA

0.8917

0.8783

NGA

ANN

0.8126

0.7942

0.7752

NGA

0.7688

NGA

–

–

–

GBM

0.8621

0.8213

0.8012

NGA

0.7245

NGA

–

–

–

Ensemble

–

–

–

–

–

–

0.8752

0.8657

NGA

  1. LS landslide, LF liquefaction, BD building damage, HT Haiti, PR Puerto Rico, HK Hokkaido, NGA no ground truth available. Bold values represent the best performance in each column.
Back to article page