Now-a-days Diabetes is an alarming issue all over the world. In Bangladesh, many individuals are affected by it. Due to overpopulation and lack of proper education, it is very difficult to provide sufficient care for diabetes patients. This paper presents a system that can detect if anyone has diabetes. A machine learning algorithm, KNN is used on a supervised dataset to detect diabetes. It also shows nearby doctor chambers through location tracking. The dataset was collected from different hospitals of Bangladesh. Due to security reason, we cannot disclose the names of the institutions from where we collected all the data. However, supervised training in such situation shows a great accuracy although no such work has found for Bangladesh region. This paper studies two different algorithms on the dataset. These are, KNN and K-means. Between them, the proposed approach achieves 99.78% accuracy, which is so far the best for detecting diabetes. We have used total number of 6219 data of different diabetes affected patients. Through this system, one can easily know if he has diabetes by giving test reports and consult with nearest certified doctors with location tracking. Resulting in saving time and money used to detect diabetes and to find preferable doctors as one do not have to go to doctors the very first time and see his health condition sitting right at their home.