Table 3 The predictive performance (AUPRC and AUROC with 95% CI) of ML models constructed using all features or only the selected features
Logistic regression | GRF | ANN | ||
|---|---|---|---|---|
All features | AUPRC | 0.079 (0.038,0.153) | 0.072 (0.034,0.145) | 0.052 (0.022, 0.120) |
AUROC | 0.681 (0.632,0.731) | 0.686 (0.642,0.730) | 0.576 (0.524,0.628) | |
Lasso features | AUPRC | 0.077 (0.038,0.151) | 0.080 (0.040, 0.155) | 0.070 (0.033,0.143) |
AUROC | 0.684 (0.636,0.733) | 0.682 (0.631,0.732) | 0.671 (0.624,0.719) | |
GRF features (72) | AUPRC | 0.075 (0.037,0.149) | 0.075 (0.036, 0.148) | 0.067 (0.031,0.139) |
AUROC | 0.679 (0.629,0.729) | 0.696 (0.651,0.741) | 0.654 (0.603,0.705) | |
TGL features | AUPRC | 0.098 (0.052,0.176) | 0.077 (0.038,0.151) | 0.076 (0.037,0.150) |
AUROC | 0.709 (0.658,0.760) | 0.701 (0.657,0.744) | 0.682 (0.634,0.730) | |
Group sum of TGL features | AUPRC | 0.091 (0.049,0.169) | 0.074 (0.036,0.148) | 0.078 (0.038,0.152) |
AUROC | 0.723 (0.678,0.768) | 0.695 (0.652,0.738) | 0.692 (0.646,0.737) | |
Group union of TGL features | AUPRC | 0.094 (0.049,0.172) | 0.074 (0.036,0.148) | 0.086 (0.044,0.163) |
AUROC | 0.726 (0.681,0.771) | 0.693 (0.647,0.738) | 0.712 (0.665,0.758) |