Research Article
Learning Air Traffic as Images: A Deep Convolutional Neural Network for Airspace Operation Complexity Evaluation
Table 3
Evaluation performance of different methods.
| Methods | Accuracy | MAE | F1 score |
| ML (features required) | GNB | 58.28% (±1.86%) | 0.4422 (±0.0228) | 47.56% (±2.78%) | KNN | 72.09% (±1.59%) | 0.2891 (±0.0260) | 62.60% (±4.42%) | LLR | 68.26% (±1.10%) | 0.3212 (±0.0118) | 46.31% (±2.72%) | SVM | 71.96% (±1.01%) | 0.2874 (±0.0113) | 52.22% (±4.14%) | MLP | 71.98% (±1.66%) | 0.2885 (±0.0163) | 59.46% (±8.19%) | RF | 69.49% (±0.93%) | 0.3129 (±0.0008) | 59.88% (±4.06%) | AdaBoost | 73.84% (±0.91%) | 0.2705 (±0.0080) | 64.16% (±9.64%) |
| DL (MTSI) | Shadow-CNN | 61.63% (±1.05%) | 0.4132 (±0.0129) | 51.23% (±5.42%) | SOCNN | 76.06% (±1.04%) | 0.2499 (±0.0080) | 70.23% (±4.61%) |
|
|