Here, we have collected database from online source, of 800 different wpeople with details of their insulin level, No of pregnancy, BMI, skin thickness, age and also the outcome whether they had diabetes or not. We made a model using machine learning to predict whether a person has diabetes or not. In this we used three different approaches Logistic Regression, Decision Tree Regeression and Random Forest Classification, after comparing the accuracy of the test data set of each approach we came to the conclusion that Random Forest Classification predicted the best and had a accuracy of 78.6% predictions.