Stroke is one of the diseases that affect humans, leading to death and disability. It occurs when some tissues in the brain die. This causes the blockage or rupture of blood vessels in the brain, which leads to damage in some of its parts. Therefore, parts of the body under the control of the damaged brain area cannot function properly. Ischemic and hemorrhagic stroke is the stroke main types. People with stroke require immediate treatment. However, the initial treatment is chosen based on the type of stroke that the patient has, whether ischemic or hemorrhagic. Therefore, it is essential to carry out stroke classification to administer appropriate initial treatment to patients. As technology grows, this can be carried out effectively and accurately using the machine learning method. In this study, the Fuzzy C-Means (FCM) with the Grey Wolf Optimization (GWO) feature selection (FCM-GWO) was implemented to classify the stroke dataset obtained from Dr. Cipto Mangunkusumo National General Hospital, Jakarta. Furthermore, the GWO feature selection was carried out on the stroke dataset to eliminate less important features and increase the performance of FCM. The results showed that the FCM-GWO and FCM achieved an accuracy value of 76% and 60%, respectively. Therefore, it was concluded that the FCM-GWO was able to increase the performance accuracy of FCM. © 2021 IEEE.