Table 3 Model performance comparison with varying forecast horizons.
From: Electric-load forecasting using interval models based on granularity and justifiable principles
Model | Lag, output | RMSE | MAE | \(R^2\) | MAPE |
|---|---|---|---|---|---|
Linear regression | lag=48, output=1 | 109.52 | 84.36 | 0.9918 | 2.00 |
Decision tree | lag=48, output=1 | 149.26 | 114.49 | 0.9847 | 2.69 |
SVM | lag=48, output=1 | 222.17 | 165.13 | 0.9661 | 3.72 |
KNN | lag=48, output=1 | 135.77 | 103.81 | 0.9873 | 2.36 |
Boosted ensemble | lag=48, output=1 | 109.95 | 84.31 | 0.9917 | 1.95 |
Neural net | lag=48, output=1 | 124.35 | 96.81 | 0.9894 | 2.34 |
LSTM | lag=48, output=1 | 109.70 | 84.11 | 0.9917 | 1.99 |
Transformer | lag=48, output=1 | 111.44 | 85.87 | 0.9915 | 2.04 |
ARIMA | lag=48, output=1 | 207.55 | 158.93 | 0.9404 | 3.69 |
SARIMA | lag=48, output=1 | 217.16 | 165.58 | 0.9676 | 3.86 |
Linear regression | lag=48, output=48 | 178.44 | 135.07 | 0.9781 | 3.14 |
Decision tree | lag=48, output=48 | 285.93 | 212.16 | 0.9438 | 4.88 |
SVM | lag=48, output=48 | 231.50 | 182.07 | 0.9631 | 4.13 |
KNN | lag=48, output=48 | 192.33 | 146.74 | 0.9746 | 3.33 |
Boosted ensemble | lag=48, output=48 | 209.43 | 156.92 | 0.9698 | 3.50 |
Neural net | lag=48, output=48 | 187.29 | 143.16 | 0.9759 | 3.36 |
LSTM | lag=48, output=48 | 233.81 | 178.62 | 0.9624 | 3.89 |
Transformer | lag=48, output=48 | 217.28 | 168.24 | 0.9675 | 3.78 |
ARIMA | lag=48, output=48 | 1766.04 | 1455.39 | -1.1449 | 29.65 |
SARIMA | lag=48, output=48 | 1758.65 | 1448.87 | -1.1270 | 29.63 |
Linear regression | lag=48, output=336 | 204.55 | 163.02 | 0.2830 | 5.66 |
Decision tree | lag=48, output=336 | 305.31 | 220.36 | 0.5973 | 7.59 |
SVM | lag=48, output=336 | 340.94 | 293.56 | -1.0030 | 10.23 |
KNN | lag=48, output=336 | 205.21 | 156.98 | 0.2780 | 5.44 |
Boosted ensemble | lag=48, output=336 | 226.87 | 176.01 | 0.1157 | 6.13 |
Neural net | lag=48, output=336 | 219.96 | 174.99 | 0.1707 | 6.06 |
LSTM | lag=48, output=336 | 332.58 | 257.87 | 0.9227 | 5.65 |
Transformer | lag=48, output=336 | 300.18 | 235.02 | 0.9370 | 5.35 |
ARIMA | lag=48, output=336 | 1745.09 | 1439.82 | -1.1247 | 28.75 |
SARIMA | lag=48, output=336 | 1713.45 | 1411.00 | -1.0484 | 28.52 |
