Table 4.
Significant predictors and accuracy for the case of General Health and different clustering techniques before and after Boruta.
| Method | Nr clusters | Accuracy (f1_macro) | Important Predictors | Accuracy (f1_macro) with important predictors |
|---|---|---|---|---|
| kml3d | 10 | 68.26 (68.02) | ‘Age’, ‘Injury severity score’, | 69.13 (68.23) |
| ‘Comorbidities’, ‘BMI’, ‘Status score’, | ||||
| ‘Pre-injury EQ-VAS’, ‘Frailty’, ‘Admission days in hospital’ | ||||
| HDclassif | 6 | 73.96 (72.59) | ‘Age’, ‘Injury severity score’, | 73.82 (72.43) |
| ‘Comorbidities’, ‘BMI’, ‘Status score’, | ||||
| ‘Pre-injury EQ-VAS’, ‘Frailty’, ‘Admission days in hospital’ | ||||
| Deepgmm | 6 | 98.20 (97.87) | ‘Age’, ‘Category accident’, ‘Admission days in hospital’, | 98.26 (97.92) |
| ‘Injury severity score’, ‘Education level’, ‘Comorbidities’, | ||||
| ‘Status score’, ‘Pre-injury EQ-VAS’, ‘Frailty’, | ||||
| ‘Traumatic brain injury’, ‘Gender’, ‘Pre-injury cognition’ |