Table 9. Feature ranking results obtained through Random Forests on the Takashi2019 diabetes type 1 dataset.
We computed the average Borda score on 1,000 executions of Random Forests. At the beginning of each execution we randomly shuffled the dataset instances.
| Rank | Feature | Average borda score | s.d. |
|---|---|---|---|
| 1 | Age | 1.275 | 1.719 |
| 2 | Bolus | 6.955 | 5.488 |
| 3 | Gait speed | 7.537 | 5.657 |
| 4 | eGFR | 8.244 | 5.683 |
| 5 | TDD | 9.683 | 5.467 |
| 6 | Grip strength | 10.517 | 5.116 |
| 7 | BMI | 10.578 | 5.319 |
| 8 | Adiponectin | 10.991 | 5.019 |
| 9 | Basal | 11.025 | 4.959 |
| 10 | HbA1c | 11.063 | 4.906 |
| 11 | Bodyfat | 11.139 | 4.664 |
| 12 | OC | 11.177 | 4.536 |
| 13 | Sex | 11.178 | 5.174 |
| 14 | Free-test | 11.200 | 4.798 |
| 15 | Knee extension strength | 11.322 | 4.777 |
| 16 | SMI | 11.458 | 4.521 |
| 17 | ucOC | 11.459 | 4.497 |
| 18 | Insulin regimen | 11.564 | 4.809 |
| 19 | Added weight | 11.635 | 4.535 |
Notes.
- s.d.
- standard deviation