Skip to main content
. 2022 Mar 10;81:103832. doi: 10.1016/j.scs.2022.103832

Table 3.

Performance of ML Models for SoE.

Model Type SoE
Overall Accuracy TP (1) TP (2) TP (3)
ANN 10-neuron 81.3% 24.2% 97.1% 63.5%
25-neuron 81.2% 26.6% 96.5% 63.0%
100-neuron 80.9% 26.6% 96.2% 62.6%
RT Coarse Tree 81.5% 22.1% 97.7% 64.1%
Medium Tree 81.5% 22.5% 97.7% 63.8%
Fine Tree 81.1% 25.9% 96.7% 61.5%
Ensembles Boosted Tree 81.4% 22.9% 97.5% 64.1%
Bagged Tree 81.2% 24.6% 97.1% 62.3%
Subspace KNN 75.0% 23.1% 90.5% 51.0%
RUBoosted Trees 79.0% 47.1% 88.5% 64.7%
SVM Linear 79.2% 0.0% 99.6% 64.1%
Quadratic 80.4% 14.4% 98.0% 64.1%
Cubic 73.9% 14.4% 88.9% 65.1%
Fine Gaussian 81.0% 22.7% 97.4% 61.5%
Medium Gaussian 81.4% 22.2% 97.8% 63.3%
Coarse Gaussian 79.2% 0.0% 99.6% 64.1%
KNN Fine KNN 72.0% 29.2% 83.5% 60.4%
Medium KNN 80.5% 28.1% 95.2% 62.8%
Coarse KNN 80.0% 8.7% 99.2% 61.0%
Cosine KNN 80.3% 27.2% 95.1% 62.9%
Cubic KNN 80.5% 28.0% 95.2% 62.8%

* TP (1): True Positive for electricity.

⁎⁎ TP (2): True Positive for natural gas.

⁎⁎⁎TP (3): True Positive for heating oil.