With the increasing usage of technology and more precisely using machine learning techniques in the medicine industries to distinguish disease types or the classification of diseases, researchers started to implement these kinds of algorithms in a variety of specific datasets to obtain better prediction or classification. The purpose of this article is to investigate the diagnosis of diabetes based on machine learning algorithms. In the study, four different machine learning models (KNN, Decision Tree, Random Forest, and Naive Bayes) were used for classification in the diagnosis of diabetes. The study focused on seven features extracted from the dataset that are HighBP, HighChol, BMI, Stroke, HDAttack, Age, and Sex. The target variable, Diabetes binary, encompasses values of 0 and 1, denoting non-diabetic and diabetic or pre-diabetic individuals respectively. The results revealed that the KNN model achieved the highest accuracy rate of 85% and an F1 score of 0.92, and had better generalization ability compared to other models. The average accuracy and F1 scores of all models were relatively high, and the models showed consistent performance on different benchmarks. In the future, this work can be expanded by adding more features, diversifying different machine-learning models, and developing hybrid methods.