Diabetes is one of the most common non-communicable diseases in the world. Diabetes affects the ability to produce the hormone insulin. Thus, complications may occur if diabetes remains untreated and unidentified. That features a significant contribution to increased morbidity, mortality, and admission rates of patients in both developed and developing countries. When disease is not detected early, it leads to complications. Medical records of the cases were retrospective. Anthropometric and biochemical information was collected. From this data, four ML classification algorithms, including Decision Tree (J48), Naive-Bayes, PART rule induction, and JRIP, were used to prognosticate diabetes. Precision, recall, F-Measure, Receiver Operating Characteristics (ROC) scores, and the confusion matrix were calculated to determine the performance of the various algorithms. The performance was also measured by sensitivity and specificity. They have high classification accuracy and are generally comparable in predicting diabetes and free diabetes patients. Among the selected algorithms tested, the Decision Tree Classifier (J48) algorithm scored the highest accuracy and was the best predictor, with a classification accuracy of 92.74%.