Research Article
Enhancing Photovoltaic Module Fault Diagnosis with Unmanned Aerial Vehicles and Deep Learning-Based Image Analysis
Table 9
Training (a), validation (b), and test (c) accuracies of tree-based classifiers for various pretrained networks.
(a) Training accuracy of tree-based classifiers for various pretrained networks |
| Classifier | Training accuracy (%) | AlexNet | DenseNet-201 | GoogleNet | ResNet-50 | VGG16 | VGG19 |
| BF tree | 99.20 | 98.76 | 98.41 | 98.41 | 98.45 | 98.13 | CS forest | 99.01 | 99.52 | 99.44 | 99.28 | 99.52 | 99.88 | Decision stump | 32.14 | 33.09 | 31.42 | 31.98 | 29.84 | 32.65 | Extra tree | 100.00 | 100.00 | 100.00 | 100.00 | 100.00 | 100.00 | Forest PA | 99.72 | 99.88 | 99.84 | 99.84 | 99.48 | 99.68 | FT | 99.92 | 99.96 | 99.76 | 99.88 | 99.48 | 99.92 | Hoeffding tree | 93.05 | 80.91 | 31.90 | 90.59 | 72.06 | 51.38 | J48 | 99.28 | 99.20 | 98.96 | 99.00 | 98.92 | 99.08 | J48graft | 99.28 | 99.20 | 98.96 | 99.00 | 98.92 | 99.08 | LAD tree | 89.16 | 91.50 | 86.74 | 89.40 | 86.26 | 89.60 | LMT | 99.56 | 100.00 | 99.96 | 99.80 | 99.92 | 99.72 | NB tree | 99.88 | 100.00 | 99.96 | 100.00 | 99.84 | 99.96 | Optimized forest | 100.00 | 100.00 | 100.00 | 100.00 | 100.00 | 100.00 | Random forest | 100.00 | 100.00 | 100.00 | 100.00 | 100.00 | 100.00 | Random tree | 100.00 | 100.00 | 100.00 | 100.00 | 100.00 | 100.00 | REP tree | 95.59 | 97.14 | 95.11 | 95.83 | 95.19 | 96.82 | Simple cart | 99.24 | 98.92 | 98.05 | 98.25 | 98.45 | 98.92 |
|
|
(b) Validation accuracy of tree-based classifiers for various pretrained networks |
| Classifier | Validation accuracy (%) | AlexNet | DenseNet-201 | GoogleNet | ResNet-50 | VGG16 | VGG19 |
| BF tree | 93.88 | 94.84 | 93.21 | 93.88 | 93.13 | 94.08 | CS forest | 95.79 | 97.30 | 95.63 | 95.55 | 96.03 | 96.66 | Decision stump | 32.10 | 33.05 | 31.42 | 31.78 | 30.07 | 32.53 | Extra tree | 91.78 | 91.58 | 91.54 | 91.70 | 91.15 | 91.34 | Forest PA | 97.93 | 98.69 | 97.26 | 98.49 | 97.69 | 97.38 | FT | 98.37 | 99.13 | 97.73 | 98.57 | 97.42 | 98.38 | Hoeffding tree | 92.50 | 91.11 | 88.49 | 90.03 | 83.92 | 89.84 | J48 | 94.44 | 95.39 | 93.61 | 93.84 | 94.20 | 94.20 | J48graft | 94.24 | 94.23 | 93.84 | 93.73 | 93.92 | 96.65 | LAD tree | 88.01 | 89.48 | 87.06 | 87.34 | 83.76 | 87.97 | LMT | 98.69 | 99.48 | 97.93 | 98.69 | 97.50 | 98.37 | NB tree | 93.01 | 93.49 | 94.66 | 92.38 | 88.41 | 89.88 | Optimized forest | 99.16 | 99.64 | 98.69 | 99.32 | 98.69 | 99.96 | Random forest | 99.08 | 99.68 | 98.76 | 99.36 | 98.73 | 99.28 | Random tree | 93.01 | 94.84 | 93.25 | 98.05 | 93.21 | 92.81 | REP tree | 92.81 | 93.45 | 91.82 | 91.36 | 91.86 | 91.78 | Simple cart | 94.28 | 95.31 | 93.45 | 93.69 | 93.25 | 94.44 |
|
|
(c) Test accuracy of tree-based classifiers for various pretrained networks |
| Classifier | Test accuracy (%) | AlexNet | DenseNet-201 | GoogleNet | ResNet-50 | VGG16 | VGG19 |
| BF tree | 95.55 | 95.23 | 93.33 | 93.96 | 94.12 | 93.17 | CS forest | 96.66 | 97.77 | 96.19 | 95.39 | 96.19 | 97.62 | Decision stump | 32.69 | 33.17 | 37.52 | 31.90 | 29.36 | 32.69 | Extra tree | 93.17 | 91.74 | 90.85 | 91.26 | 92.69 | 92.53 | Forest PA | 98.09 | 98.57 | 97.71 | 98.73 | 97.69 | 97.93 | FT | 98.41 | 99.2 | 97.90 | 99.04 | 97.3 | 98.41 | Hoeffding tree | 93.01 | 80.31 | 19.42 | 89.84 | 70.79 | 49.68 | J48 | 95.23 | 95.07 | 94.66 | 94.13 | 94.76 | 94.44 | J48graft | 95.71 | 95.23 | 95.80 | 93.96 | 93.96 | 94.28 | LAD tree | 89.52 | 90.15 | 84.57 | 86.98 | 83.33 | 86.82 | LMT | 99.20 | 99.74 | 98.28 | 99.36 | 98.25 | 98.41 | NB tree | 94.44 | 95.55 | 95.42 | 96.19 | 93.17 | 93.33 | Optimized forest | 99.04 | 99.80 | 99.03 | 99.68 | 99.36 | 99.80 | Random forest | 99.52 | 99.84 | 99.23 | 99.68 | 99.68 | 99.84 | Random tree | 94.44 | 94.12 | 90.47 | 90.31 | 93.53 | 94.60 | REP tree | 92.69 | 95.39 | 90.85 | 91.74 | 92.16 | 93.49 | Simple cart | 95.55 | 95.07 | 93.33 | 94.12 | 94.28 | 94.12 |
|
|
Boldface entries represent the highest values obtained.
|