From: Procedure code overutilization detection from healthcare claims using unsupervised deep learning methods
Test Datasets
Metrics
DBSCAN
Autoencoder
Out-of-sample
Precision
0.22
0.87
Recall
0.58
1.0
F1
0.32
0.93
Manually annotated
0.19
0.36
0.41
0.86
0.26
0.51