Diabetes is a chronic illness that can have deadly consequences. It has emerged as a worldwide health emergency. For prompt management and the prevention of disease development, early identification is essential. Using a range of Machine Learning (ML) methods, such as Support Vector Machine (SVM), k-Nearest Neighbours (KNN), Random Forest (RF), XGBoost, Convolutional Neural Networks (CNN), and ResNet, this study suggests a thorough method for diabetes prediction. These models were tested on two different datasets to determine how well they predict the onset of diabetes. Several performance indicators, including accuracy, precision, recall, F1 score, and area under the receiver operating characteristic curve, were used to train the models. Showing a test accuracy of 0.88 and an F1 score of 0.6186 on first dataset, RF showed the best generalization ability and accuracy of all the studied models. Additionally, SVM and XGBoost fared well, especially when it came to lowering false positives and striking a balance between recall and specificity, respectively. Nevertheless, overfitting was evident in deep learning models such as CNN and ResNet, suggesting that cautious cross-validation is necessary.