Table 4.
RMSE and MAE values for multivariable models in five different forecasting horizons.
| Models | Forecasting Horizon (in days) | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 7 | 14 | 21 | 30 | ||||||
| RMSE | MAE | RMSE | MAE | RMSE | MAE | RMSE | MAE | RMSE | MAE | |
| SARIMAX | 35.57 | 31.08 | 39.76 | 34.75 | 48.27 | 43.02 | 52.89 | 46.96 | 60.92 | 53.91 |
| FP | 48.68 | 43.25 | 58.27 | 53.37 | 70.07 | 65.56 | 80.34 | 76.13 | 78.00 | 72.39 |
| LSTM | 28.55 | 20.52 | 30.04 | 21.32 | 31.26 | 22.14 | 31.20 | 23.54 | 35.96 | 28.03 |
Note: Green highlights the model with the best results. Interpretation: The LSTM model had the smallest RMSE and MAE values for the 30-day forecasting horizon. MAE, mean absolute error; RMSE, root mean square error (units = patients).