Table 6.
Overall importance ranking of each risk factor in predicting bleeding based on RF and XGBoost.
| Risk Factors | Average Ranking of 10 Times RF | Average Ranking of 10 Times XGBoost | Average Ranking of the 2 Models | Final Ranking in Predicting Bleeding |
|---|---|---|---|---|
| Age | 1 | 1.3 | 1.15 | 1 |
| Kidney function | 3.2 | 3.5 | 3.35 | 2 |
| Smoking | 2.1 | 4.7 | 3.4 | 3 |
| Previous bleeding history | 4.7 | 2.4 | 3.55 | 4 |
| Concomitant use of drugs | 4.8 | 5 | 4.9 | 5 |
| Medicine dosage (dabigatran) | 7 | 6.7 | 6.85 | 6 |
| BMI | 5.2 | 9.6 | 7.4 | 7 |
| History of myocardial infarction | 9.2 | 6.1 | 7.65 | 8 |
| History of congestive heart failure | 10.1 | 10.3 | 10.2 | 9 |
| Ethnicity | 9.1 | 12.2 | 10.65 | 10 |
| Sex | 10.8 | 11.3 | 10.55 | 11 |
| History of diabetes mellitus | 12.8 | 11.5 | 12.15 | 12 |
| Previous stroke history | 11 | 14.2 | 12.6 | 13 |
| Hypertension history | 14 | 12.6 | 13.3 | 14 |
| Body weight | 15 | 12.3 | 13.65 | 15 |
| History of systemic embolism | 16 | 13 | 14.5 | 16 |
| Liver function abnormality | 17 | 17 | 17 | 17 |
| Anemia | 18 | 17.4 | 17.7 | 18 |
Abbr.: RF, random forest; XGBoost, eXtreme gradient boosting; BMI, body mass index.