Gestational Diabetes Mellitus (GDM) is widely acknowledged as a significant and common health concern among pregnant women, posing serious risks if not diagnosed early. Traditionally, diagnosing GDM requires several visits to healthcare facilities and consultations with various specialists, making the process both complex and time-consuming. Fortunately, machine learning methods present a promising solution to address these challenges. The goal of this study is to create a model that can reliably predict the likelihood of GDM in patients. To accomplish this, three machine learning (ML) classifiers—K-Nearest Neighbors (K-NN), Decision Tree (DT), and Extreme Gradient Boosting (XGBoost)—were utilized to identify GDM at an early stage. A series of experiments were carried out using a dataset gathered from Benghazi Medical Center and several Libyan clinics, with support from doctors. The dataset was imbalanced, and to address this issue, various oversampling techniques were applied. The performance of the selected classifiers was assessed using standard metrics, including precision, accuracy, F1-score, and recall. The experimental results revealed that the XGBoost tree model delivered the best performance, achieving an AUC value of 87%.