Table 5 Sensitivity analysis 3 for identifying the appropriate learning data size.
From: Developing a hybrid time-series artificial intelligence model to forecast energy use in buildings
Scenario | Learning data size (months) | Accuracy indices with the test data | ANC | Rank | |||
|---|---|---|---|---|---|---|---|
RMSE (kWh) | MAE (kWh) | MAPE (%) | R | ||||
1 | 3 | 1.23 | 1.52 | 4.65 | 0.991 | 0.485 | 8 |
2 | 4 | 1.21 | 1.46 | 4.40 | 0.992 | 0.091 | 1 |
3 | 5 | 1.23 | 1.52 | 4.59 | 0.991 | 0.296 | 3 |
4 | 6 | 1.24 | 1.54 | 4.68 | 0.991 | 0.389 | 7 |
5 | 7 | 1.24 | 1.53 | 4.64 | 0.991 | 0.369 | 6 |
6 | 8 | 1.26 | 1.59 | 4.75 | 0.991 | 0.557 | 10 |
7 | 9 | 1.27 | 1.62 | 4.83 | 0.990 | 0.730 | 12 |
8 | 10 | 1.28 | 1.64 | 4.89 | 0.990 | 0.819 | 13 |
9 | 11 | 1.25 | 1.56 | 4.79 | 0.991 | 0.629 | 11 |
10 | 12 | 1.25 | 1.55 | 4.72 | 0.991 | 0.494 | 9 |
11 | 13 | 1.23 | 1.50 | 4.57 | 0.991 | 0.326 | 4 |
12 | 14 | 1.22 | 1.50 | 4.54 | 0.991 | 0.333 | 5 |
13 | 15 | 1.20 | 1.45 | 4.50 | 0.992 | 0.242 | 2 |
