Synthetic Temporal Dataset for Early Prediction of Gestational Diabetes Mellitus

This dataset is developed to support robust early prediction of gestational diabetes mellitus (GDM). Due to limited access to large clinical cohorts, a publicly available GDM dataset from Kaggle was adopted to ensure transparency and reproducibility. The raw data were cleaned and class imbalance was addressed using SMOTE. The resulting balanced static dataset was then expanded into week-wise temporal sequences using a physiology-guided augmentation approach. These synthetic trajectories were validated to align with established maternal physiological trends, forming the basis for subsequent model development.