Within-person depression trajectory classification on CES-D severity crossing label N = 411 (test)
84.2Balanced AccuracyLightGBM
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| LightGBMSelection criterion=Best Accuracy and Sensitivity, Model category=Machine learning model2026.04 | 84.2 | 90.1 | 86.5 | 34.4 | |
| XGBoostSelection criterion=Best AUC, Model category=Machine learning model2026.04 | 83.4 | 90.6 | 83.8 | 35.6 | |
| ElasticNetModel category=Machine learning model2026.04 | 69.9 | 83.4 | 73 | 24.5 | |
| SVMModel category=Machine learning model2026.04 | 69.6 | 84.1 | 64.9 | 30.4 | |
| Regression to person's meanModel category=Rule-based baseline, Note=Strongest baseline2026.04 | 67.4 | 75 | 54.1 | 40.8 | |
| Last value carried forwardModel category=Rule-based baseline2026.04 | 33.6 | 55.7 | 5.4 | 5.3 | |
| Predict all stableModel category=Rule-based baseline2026.04 | 33.3 | 50 | 0 | — | |
| Person-specific modal classModel category=Rule-based baseline2026.04 | 32.7 | 49.9 | 0 | — |