Table 11.
Performance of the length of stay model results based on the area under the receiver operating characteristic curve (AUROC).
| Method and algorithm | Training set AUROC, mean (SD) | Test set AUROC | |
| Baseline approach |
|
|
|
|
|
LRa | 0.612883 (0.006047) | 0.60833 |
|
|
LDAb | 0.612776 (0.006058) | 0.60837 |
|
|
RFc | 0.664959 (0.006147) | 0.66325 |
|
|
kNNd | 0.583710 (0.006401) | 0.59110 |
|
|
SVMe | 0.665992 (0.006041) | 0.66118 |
|
|
XGBf | 0.677454 (0.007311) | 0.66586 |
| Quantiles approach |
|
|
|
|
|
LR | 0.654390 (0.012180) | 0.65407 |
|
|
LDA | 0.654178 (0.012102) | 0.65384 |
|
|
RF | 0.705115 (0.010004) | 0.69782 |
|
|
kNN | 0.598228 (0.007539) | 0.60507 |
|
|
SVM | 0.694473 (0.009834) | 0.69272 |
|
|
XGB | 0.704889 (0.011338) | 0.69693 |
aLR: logistic regression.
bLDA: linear discriminant analysis.
cRF: random forest.
dkNN: k-nearest neighbor.
eSVM: support vector machine.
fXGB: extreme gradient boosting.