Skip to main content

Table 6 Stratified customized loss functions (MAE) of one-stage and two-stage models with 95% bootstrap confidence intervals

From: Predicting in-hospital length of stay: a two-stage modeling approach to account for highly skewed data

 

One-stage model (log data)

Two-stage model (log data)

LASSO

RF

MLP

0–2 days

0.744 (0.739, 0.749)

0.736 (0.731, 0.741)

1.596 (1.585, 1.607)

1.313 (1.305, 1.321)

1.638 (1.626, 1.648)

2–4 days

0.713 (0.709, 0.718)

0.705 (0.700, 0.709)

1.046 (1.037, 1.058)

1.035 (1.027, 1.047)

1.155 (1.142, 1.168)

4–7 days

1.782 (1.771, 1.794)

1.760 (1.750, 1.772)

1.492 (1.474, 1.509)

1.586 (1.571, 1.600)

1.730 (1.713, 1.746)

0–7 days

0.927 (0.922, 0.930)

0.915 (0.911, 0.919)

1.358 (1.351, 1.366)

1.254 (1.248, 1.261)

1.464 (1.456, 1.471)