Table B4.

Comparison between hybrid models derived using equal weights and weighted errors considering minimization of MAE, MAPE and RMSE (patients hospitalized in the ICU)

Model	MAE (%)	MAPE (%)	RMSE (%)
ARIMA–ETS	− 0.72	− 0.28	− 1.04
ARIMA–NNAR	− 1.87	− 1.16	− 2.87
ARIMA–TBATS	− 1.74	− 0.17	− 1.83
ETS–NNAR	− 4.01	− 1.54	− 5.56
ETS–TBATS	− 0.01	−	0.01
NNAR–TBATS	− 6.5	− 0.77	− 8.57
ARIMA–ETS–NNAR	− 2.95	− 0.91	− 4.25
ARIMA–ETS–TBATS	− 3.23	− 3.47	− 3.65
ARIMA–NNAR–TBATS	− 3.69	− 0.33	− 5.07
ETS–NNAR–TBATS	− 3.13	− 0.75	− 4.39
ARIMA–ETS–NNAR–TBATS	− 3.42	− 0.98	− 4.38

Notes: roman values indicated that equal weights were better, while italic colored values indicated that weighted errors were better.