Table 8 Performance of different boosting algorithms under \(CV_3.\)
From: Predicting lncRNA and disease associations with graph autoencoder and noise robust gradient boosting
| Â | Dataset | XGBoost | AdaBoost | CatBoost | LightGBM | LDA-GARB |
|---|---|---|---|---|---|---|
Precision | Dataset 1 | 0.8624 ± 0.0222 | 0.8085 ± 0.0251 | 0.8737 ± 0.0285 | 0.8531 ± 0.0267 | 0.8743\({\varvec{\pm }}\)0.0284 |
Dataset 2 | 0.8983 ± 0.0156 | 0.9041 ± 0.0155 | 0.9121 ± 0.0173 | 0.9070 ± 0.0130 | 0.9348\({\varvec{\pm }}\)0.0130 | |
Recall | Dataset 1 | 0.8450 ± 0.0334 | 0.8147 ± 0.0338 | 0.8590 ± 0.0320 | 0.8245 ± 0.0303 | 0.8724\({\varvec{\pm }}\)0.0305 |
Dataset 2 | 0.9052 ± 0.0144 | 0.8596 ± 0.0193 | 0.9073 ± 0.0144 | 0.9115 ± 0.0148 | 0.9373\({\varvec{\pm }}\)0.0137 | |
Accuracy | Dataset 1 | 0.8547 ± 0.0178 | 0.8103 ± 0.0197 | 0.8668 ± 0.0189 | 0.8408 ± 0.0201 | 0.8729\({\varvec{\pm }}\)0.0204 |
Dataset 2 | 0.9012 ± 0.0118 | 0.8841 ± 0.0129 | 0.9097 ± 0.0104 | 0.9089 ± 0.0089 | 0.9359\({\varvec{\pm }}\)0.0085 | |
F1-score | Dataset 1 | 0.8531 ± 0.0192 | 0.8110 ± 0.0207 | 0.8657 ± 0.0192 | 0.8381 ± 0.0208 | 0.8728\({\varvec{\pm }}\)0.0204 |
Dataset 2 | 0.9016 ± 0.0117 | 0.8811 ± 0.0135 | 0.9096 ± 0.0102 | 0.9091 ± 0.0090 | 0.9359\({\varvec{\pm }}\)0.0085 | |
AUC | Dataset 1 | 0.9260 ± 0.0131 | 0.8788 ± 0.0168 | 0.9427 ± 0.0109 | 0.9163 ± 0.0141 | 0.9459\({\varvec{\pm }}\)0.0109 |
Dataset 2 | 0.9679 ± 0.0050 | 0.9580 ± 0.0065 | 0.9723 ± 0.0045 | 0.9689 ± 0.0053 | 0.9790\({\varvec{\pm }}\)0.0051 | |
AUPR | Dataset 1 | 0.9240 ± 0.0145 | 0.8750 ± 0.0194 | 0.9411 ± 0.0108 | 0.9135 ± 0.0174 | 0.9418\({\varvec{\pm }}\)0.0136 |
Dataset 2 | 0.9690 ± 0.0048 | 0.9601 ± 0.0065 | 0.9732 ± 0.0046 | 0.9696 ± 0.0061 | 0.9744\({\varvec{\pm }}\)0.0100 |
