Predicting Gestational Diabetes Mellitus Using Machine Learning Algorithms: A Comprehensive Review and Analysis

Abstract

Predicting pregnancy-related diabetes using machine learning (ML) provides a proactive healthcare strategy and permits identifying at-risk patients well in advance. This helps healthcare practitioners to consider early treatment options that can mitigate the risk of complications for mother and child. It examines large datasets for subtle patterns and underlying risk variables, providing better predictive accuracy. By managing gestational diabetes early, medical providers mitigate the risks of macrosomia, preterm birth, and maternal hypertension. Finally, the approach leads to more targeted and effective prenatal care, improving health outcomes and securing a better start for the baby and mother. Besides, predictions of diabetes in pregnancy are supposed to be carried out by using RF-a ML classification algorithm. Red Deer Algorithm (RDA), Jellyfish Search Optimizer (JSO), and Rhizotomy Optimization Algorithm (ROA) are utilized to increase the model's precision. This model has been selected to integrate with optimizers to enhance forecast accuracy. In diabetes circumstances, RFJS is observed to outperform other models, with a precision value of 0.9377, while the rank of the second best is from the RFRO model, having a precision of 0.932. The RFRD scheme has a precision of 0.9282, thus reflecting medium performance but higher than the RFC with its precision of 0.879. These results have provided significant insights into the current performance and potential of ML algorithm models in terms of prediction. The insight of such a result point out the direction of future research and further development in applying ML to predictive analytics in diverse fields.