Diabetes mellitus is a perdurable hyperglycemic disease. Various complications can be caused by this disease. In line with the growing morbidness in the last few years, 642 million people can be infected with diabetes within 2040 which is one among 10 individuals. So undoubtedly this malady needs more attention. Nowadays the usage of machine learning is increasing. So, in many medical perspectives, this technique has been utilized. We have chosen methodologies that give the best performances for independent testing to confirm the universal applicability of the techniques. We have focused on early detecting this disease. We have collected data from Khulna Diabetes Center at Khulna where the instances is 289 and 13 features. In our study, we use the Logistic Regression model with 88%, XGboost 86.36% and Random Forest with 86.36% accuracy. We found that the random forest model performs the best output for diabetics detection.