Research Article
Avionics Graphics Hardware Performance Prediction with Machine Learning
Table 4
Central tendencies of the prediction error distributions for each machine learning algorithm and for each validation dataset.
| Validation dataset name | Supervised learning algorithm | RMSE (FPS) | Relative prediction error (%) | Mean absolute prediction error (FPS) | Median absolute prediction error (FPS) |
| #1 | Random forest | 2.71 | 2.12 | 1.36 | 0.45 | MART | 9.37 | 10.15 | 6.85 | 4.86 | Neural net | 2.29 | 1.99 | 1.26 | 0.50 |
| #2 | Random forest | 10.87 | 8.74 | 5.85 | 2.06 | MART | 16.53 | 17.65 | 12.16 | 10.39 | Neural net | 7.90 | 8.50 | 5.16 | 3.51 |
| #3 | Random forest | 5.94 | 5.98 | 4.10 | 2.79 | MART | 9.99 | 10.60 | 7.51 | 6.15 | Neural net | 2.97 | 3.03 | 2.39 | 2.13 |
| #4 | Random forest | 15.80 | 9.12 | 10.94 | 3.98 | MART | 30.20 | 18.78 | 22.53 | 15.02 | Neural net | 4.02 | 2.58 | 3.10 | 3.01 |
|
|
Figures 6ā 9 present the error distribution of each nonparametric performance model for the four validation scenes. |