Table 1.
Comparison of prediction performance using different predictive models.
| Feature method | Network-based | Edge-based | |||
|---|---|---|---|---|---|
| Regression model | NKI-RS-E (K-fold) | NKI-RS (external) | NKI-RS-E (K-fold) | NKI-RS (external) | |
| OLS | Pearson's correlation | 0.846 (p < 0.0001) |
0.786 (p < 0.0001) |
0.867 (p < 0.0001) |
0.828 (p < 0.0001) |
| MAE (years) | 7.9 | 9.8 | 6.7 | 9.3 | |
| SVR | Pearson's correlation | 0.868 (p < 0.0001) |
0.769 (p < 0.0001) |
0.890 (p < 0.0001) |
0.832 (p < 0.0001) |
| MAE (years) | 7.5 | 9.9 | 7.1 | 9.4 | |
| LASSO | Pearson's correlation | 0.910 (p < 0.0001) |
0.838 (p < 0.0001) |
0.896 (p < 0.0001) |
0.835 (p < 0.0001) |
| MAE (years) | 6.5 | 8.8 | 6.9 | 9.2 | |
Feature selection methods: network-based method and edge-based method; regression model: ordinary linear regression (OLS method), SVR and Lasso.