The World Health Organization (WHO) predicts that 15% of the world’s population will have at least one disability by 2050, with stroke being the most common. Around the world, 15 million people experience a stroke each year. Five million of them pass away, and another five million suffer from lasting disabilities. When blood vessels burst or arteries get blocked, the brain’s blood supply is cut off, resulting in damage that’s called a stroke. Early stroke identification is now quite likely thanks to the development of machine learning in prognosis. In this study, data imputation is done in several ways to analyze which method is reliable and works fine by using publicly accessible datasets, and a data balancing technique is also adapted. Here, imputation is done by filling the missing value with zero, by single imputation, or by dropping the value, and data balancing is accomplished using the SMOTE technique. Before performing SMOTE, the size of the majority and minority classes was 3648, 184, and after applying SMOTE, the ratio was balanced at 3648 and 3648. In this work, ML algorithms such as random forest, logistic regression, and decision tree are used to forecast brain attacks. Only minor changes were observed among the classifiers with random forest, which achieved 96%.