Overview of Machine and Deep Learning Models for Gestational Diabetes Mellitus (GDM) Prediction

Abstract

Gestational Diabetes Mellitus (GDM) is a form of diabetes that develops during pregnancy, posing significant risks to both mother and child. Although conventional diagnostic tests are reliable, they are often time-consuming and invasive. To address these limitations, recent studies have increasingly turned to data-driven predictive methods. Machine Learning (ML) algorithms, including Decision Trees, Random Forests, and XGBoost, have demonstrated strong potential in early GDM identification. Furthermore, advanced Deep Learning (DL) models like Recurrent Neural Networks (RNN) and Long Short-Term Memory (LSTM) have shown superior capability in modelling complex temporal health data. Hybrid architectures, such as GRU-LSTM, have reported predictive accuracies as high as 99%. Despite these promising outcomes, challenges persist in addressing class imbalance and effective data preprocessing. Comparative evaluations highlight trade-offs between predictive performance, interpretability, and computational efficiency. Future research should focus on leveraging diverse datasets, integrating explainable AI (XAI) methods, and enhancing clinical integration to support the real-world deployment of robust GDM prediction systems.