Table 4.
Top five most important predictors.
| Algorithm: | Logistic | SVM | RF | XGBoost |
|---|---|---|---|---|
| Panel A: suicidal ideation (N = 3292) | ||||
| Feature 1 | CESD score | CESD score | CESD score | CESD score |
| Feature 2 | Self-esteem | Self-esteem | Self-esteem | Self-esteem |
| Feature 3 | Satisfaction with family relation | Income | Income | Income |
| Feature 4 | Smoking | Life satisfaction | Consumption | Satisfaction with family relation |
| Feature 5 | Unpaid rent | Consumption | Net worth | Life satisfaction |
| Panel B: suicide planning or attempt (N = 488) | ||||
| Feature 1 | CESD score | CESD score | CESD score | CESD score |
| Feature 2 | Mother's education | Self-esteem | Self-esteem | Self-esteem |
| Feature 3 | Smoking | Income | Income | Life satisfaction |
| Feature 4 | Religion | Life satisfaction | Consumption | Income |
| Feature 5 | Age | Consumption | Life satisfaction | Smoking |
Notes: SVM, support vector machine; RF, random forest; XGBoost, extreme gradient boosting.