Stroke is one of the most common and dangerous diseases around the world, as this disease require an accurate and early diagnosis to reduce its severity and receive appropriate treatment in a timely manner to preserve the patient’s life. In this paper, an efficient methodology is proposed to predict stroke based on the stroke dataset obtained from Kaggle. This methodology consisted of three sequential stages to obtain the final prediction of the disease, whether the patient will have a stroke or not. Firstly, we performed data pre-processing to address challenges such as missing values by using the mean and data imbalance by employing the SMOTE technique to ensure unbiased training. Secondly, we divided the dataset into two parts: a training part and a testing part. The data were passed to the models used. Lastly, we applied five machine learning algorithms to test their performance, and the two best models were passed in terms of performance to our proposed model, which was the hard voting classifier to enhance prediction accuracy. Our results showed that hard voting was superior to the rest of the models, as well as over a set of previous studies, where it achieved an accuracy of 97.84%. To ensure the robustness of our approach, we also employed K-fold cross-validation (k = 10), achieving a performance score of 97.1%.