A stroke is a medicinal exigency and requires early prognosis in accordance with damage to the brain from intrusion of its blood circulation, therefore, early diagnosis helps, medical health professionals to save human lives. This aim can be achieved using the various machine learning techniques. In this research article, machine learning models are deployed on well known heart stroke classification data-set. In addition, effect of well established feature selection technique also observed on aforementioned machine learning models. In the experimental analysis, machine learning models with standard feature selection technique are tested on the data-set, namely, framingham, and obtained results are evaluated using the confusion metrics including recall, F1-score, precision and accuracy. From the obtained results, it is observed that Random Forest (RF) and Extra Trees (ET) performed the best with PCA (Principle component analysis), giving the highest accuracy of 88.91 %.