Table 2.
Summary of the model performance using five-fold cross validation. Data are presented as mean (range).
| Model | Training accuracy |
Validation accuracy |
AUROC | F1 score |
|---|---|---|---|---|
| Logistic Regression* | 0.757 (0.726-0.773) |
0.716 (0.643-0.762) |
0.769 (0.741-0.815) |
0.729 (0.662-0.773) |
| Random Forest | 0.963 (0.958-0.967) |
0.740 (0.699-0.764) |
0.706 (0.676-0.732) |
0.732 (0.688-0.751) |
| LASSO Regression | 0.786 (0.773-0.794) |
0.785 (0.748-0.826) |
0.781 (0.736-0.822) |
0.762 (721-0.802) |
| DeepLIFT | 0.962 (0.940-0.972) |
0.751 (0.706-0.812) |
0.716 (0.680-0.753) |
0.748 (0.721-0.760) |
Notes: AUROC, area under receiver operating characteristic curve; DeepLIFT, Deep Learning Important FeaTures; LASSO, Least Absolute Shrinkage and Selection Operator.
*
The logistic regression values are derived from a prior publication.31