Table 1.
Previously published machine learning-based GDM risk prediction models.
| Authors | Subjects/data | Algorithm | Specificity | Sensitivity | AUC-ROC |
|---|---|---|---|---|---|
| Qiu et al.17 | 4,378 women | Cost-sensitive hybrid model of logistic regression, support vector machine and CHAID tree | 0.998 | 0.622 | 0.847 |
| Zheng et al.18 | 4,771 women | Multivariate Bayesian logistic regression | 0.75 | 0.66 | 0.766 |
| Ye et al.19 | 22,242 pregnancies | Gradient boosting decision tree | 0.99 | 0.15 | 0.74 |
| 0.26 | 0.90 | ||||
| Artzi et al.20 | 588,622 pregnancies | Gradient boosting | – | – | 0.854 |
| Xiong et al.21 | 490 women | Light gradient boosting machine | 0.995 | 0.883 | 0.942 |
| Yan et al.22 | 3,988 women | Logistic regression | – | 0.706 | 0.779 |
| Hou et al.23 | 1,000 samples | Light gradient boosting machine | – | – | 0.852 |
| Wu et al.24 | 32,190 women | Deep neural network | 0.82 | 0.63 | 0.80 |
| Wu et al.25 | 17,005 women | Random forest | 0.269 | 0.91 | 0.746 |
| 0.524 | 0.75 | ||||
| 0.777 | 0.487 |
Where multiple models were developed, the best performing model is described.