Table 21 Performance when using different feature selection methods under \(CV_3\).
From: Predicting lncRNA and disease associations with graph autoencoder and noise robust gradient boosting
| Â | Dataset | Linear feature | Nonlinear feature | LDA-GARB |
|---|---|---|---|---|
Precision | Dataset 1 | 0.8714 ± 0.0283 | 0.8361 ± 0.0271 | 0.8743 ± 0.0284 |
Dataset 2 | 0.9328 ± 0.0119 | 0.9082 ± 0.0177 | 0.9348 ± 0.0130 | |
Recall | Dataset 1 | 0.8731 ± 0.0301 | 0.8555 ± 0.0286 | 0.8724 ± 0.0305 |
Dataset 2 | 0.9330 ± 0.0149 | 0.8994 ± 0.0180 | 0.9373 ± 0.0137 | |
Accuracy | Dataset 1 | 0.8717 ± 0.0227 | 0.8433 ± 0.0197 | 0.8729 ± 0.0204 |
Dataset 2 | 0.9328 ± 0.0095 | 0.9041 ± 0.0121 | 0.9359 ± 0.0085 | |
F1-score | Dataset 1 | 0.8719 ± 0.0227 | 0.8452 ± 0.0193 | 0.8728 ± 0.0204 |
Dataset 2 | 0.9328 ± 0.0097 | 0.9036 ± 0.0122 | 0.9359 ± 0.0085 | |
AUC | Dataset 1 | 0.9315 ± 0.0159 | 0.9256 ± 0.0143 | 0.9459 ± 0.0109 |
Dataset 2 | 0.9782 ± 0.0049 | 0.9710 ± 0.0057 | 0.9790 ± 0.0051 | |
AUPR | Dataset 1 | 0.9257 ± 0.0224 | 0.9186 ± 0.0204 | 0.9418 ± 0.0136 |
Dataset 2 | 0.9762 ± 0.0063 | 0.9712 ± 0.0061 | 0.9744 ± 0.0100 |
