Worldwide, millions of people struggle with the chronic illness of diabetes. Early detection and treatment of diabetes can prevent or delay complications. Both diabetes diagnosis and prediction have greatly benefited from the use of ML techniques. With the use of a dataset of clinical and demographic characteristics, we present in this work a ML-based method for diabetes prediction. Comparisons are made between Decision Tree, RF, Naive Bayes, Boosting Algorithm, and SVM in terms of their efficacy. Our findings show that the Naive Algorithm outperformed the other algorithms, with an accuracy of 76%. The most significant indicators of diabetes are, according to our research, pregnancies, BP, skin thickness, insulin, a family history of diabetes, age, glucose levels and Body Mass Index (BMI). Our method may be used as a screening tool for early diabetes identification, enabling prompt management and intervention.