The situation when the blood circulation of some areas of brain cut of is known as brain stroke. Nowadays, it is a very common disease and the number of patients who attack by brain stroke is skyrocketed. There have lots of reasons for brain stroke, for instance, unusual blood circulation across the brain. It's much more monumental to diagnostic the brain stroke or not for doctor, but the main problem of them is to accurately identify the diseases. Most of the case, doctors first observe the test result and then give their decision it is brain stroke or not. This procedure, however, is very time consuming and do not give guaranty the result is 100% accurate all the time. Some cases, doctor make some fast decision by themselves without test result and they are failed to make accurate judgement. In order to solve above problem, we need to develop a system so that we can instantly and accurately detect brain stroke (early brain stroke). In this research work, we developed a system through which we can predict brain stroke earlier and very firstly. This system used some classification algorithms such as Logistic Regression, Classification and Regression Tree, K-Nearest Neighbor and Support Vector Machine for train this model. The highest accurate is provided by KNN algorithm and the accuracy is 97%. Our proposed system is very reliable, automatic and time saving approach.