Table 5.
Importance of the variable scaled according the method varImp in the caret library for the complete data set.
| Important variables | Overall | GBM | RF | XGBoost |
|---|---|---|---|---|
| Memory complaint | 14.31 | 9.69 | 3.93 | 19.95 |
| Overnight sleeping time | 11.93 | 9.04 | 6.82 | 14.87 |
| Educational attainment (linear contrast) | 11.29 | 8.85 | 5.51 | 14.41 |
| Reading (linear contrast) | 10.73 | 9.75 | 5.35 | 13.33 |
| Age | 6.86 | 9.00 | 6.76 | 6.41 |
| Sex | 4.41 | 4.84 | 1.27 | 5.69 |
| N06 | 4.12 | 3.14 | 1.38 | 5.55 |
| M01 | 3.87 | 1.27 | 0.63 | 5.90 |
| N06B | 2.71 | 3.01 | 0.82 | 3.47 |
| Job | 2.40 | 1.74 | 2.00 | 2.72 |
| Sleeping time | 2.24 | 5.90 | 5.38 | 0.00 |
| Games (linear contrast) | 1.52 | 1.76 | 1.46 | 1.49 |
| M01A | 1.40 | 1.74 | 0.49 | 1.73 |
| Smoking (linear contrast) | 1.39 | 0.33 | 0.44 | 2.06 |
| M02 | 1.02 | 0.31 | 0.39 | 1.47 |
| C08 | 0.83 | 1.79 | 1.28 | 0.40 |
| TV consumption | 0.79 | 1.30 | 2.30 | 0.00 |
| Day time nap | 0.78 | 1.68 | 2.07 | 0.00 |
| Educational attainment (cubic contrast) | 0.64 | 2.94 | 0.85 | 0.00 |
| Exercises (cubic contrast) | 0.57 | 0.84 | 1.63 | 0.04 |
Variables have been ordered depending on the importance in the ensemble model.