Table 3 Mean and standard deviation of prediction performance for heat pump existence with different machine learning algorithms (Data: ISO week 10, 2020)
From: Detection of heat pumps from smart meter and open data
Algo. | Models | AUC | Precision ∗ | Recall | F1 |
|---|---|---|---|---|---|
RF | Model 1 | 0.794 (0.10) | 0.733 (0.17) | 0.401 (0.19) | 0.691 (0.23) |
RF | Model 2 | 0.797 (0.11) | 0.728 (0.17) | 0.412 (0.19) | 0.689 (0.23) |
RF | Model 3 | 0.822 (0.07) | 0.737 (0.21) | 0.449 (0.13) | 0.723 (0.21) |
RF | Model 4 | 0.807 (0.07) | 0.750 (0.22) | 0.426 (0.12) | 0.715 (0.21) |
SVM | Model 1 | 0.769 (0.11) | 0.822 (0.16) | 0.433 (0.19) | 0.724 (0.22) |
SVM | Model 2 | 0.773 (0.11) | 0.847 (0.16) | 0.443 (0.18) | 0.733 (0.21) |
SVM | Model 3 | 0.792 (0.10) | 0.803 (0.23) | 0.449 (0.15) | 0.736 (0.21) |
SVM | Model 4 | 0.786 (0.11) | 0.790 (0.23) | 0.449 (0.15) | 0.734 (0.21) |
kNN | Model 1 | 0.637 (0.06) | 0.524 (0.20) | 0.420 (0.13) | 0.641 (0.22) |
kNN | Model 2 | 0.663 (0.12) | 0.447 (0.22) | 0.481 (0.26) | 0.650 (0.26) |
kNN | Model 3 | 0.641 (0.09) | 0.486 (0.13) | 0.426 (0.18) | 0.639 (0.23) |
kNN | Model 4 | 0.605 (0.11) | 0.374 (0.20) | 0.394 (0.19) | 0.599 (0.26) |
NB | Model 1 | 0.689 (0.12) | 0.267 (0.05) | 0.830 (0.08) | 0.413 (0.11) |
NB | Model 2 | 0.693 (0.11) | 0.263 (0.05) | 0.831 (0.08) | 0.393 (0.12) |
NB | Model 3 | 0.701 (0.10) | 0.272 (0.04) | 0.821 (0.14) | 0.446 (0.09) |
NB | Model 4 | 0.704 (0.11) | 0.278 (0.04) | 0.843 (0.13) | 0.452 (0.09) |
ANN | Model 1 | 0.529 (0.09) | - | 0.030 (0.09) | 0.456 (0.43) |
ANN | Model 2 | 0.545 (0.05) | - | 0.033 (0.10) | 0.455 (0.43) |
ANN | Model 3 | 0.500 (0.02) | - | 0.000 (0.00) | 0.435 (0.45) |
ANN | Model 4 | 0.536 (0.05) | - | 0.000 (0.00) | 0.435 (0.45) |