Using Machine Learning Method to Predict Stroke Risk

To prevent the potential incidence of the stroke and avoid any side effects which further leads to the stroke, it becomes essential to predict whether the individual is at high risk of stroke based on his/her basic information. Previously published most of the studies in stroke prediction used Cox proportional hazards model or other novel integrated models. These models are difficult for individuals who probably do not possess sufficient background knowledge to implement and use those models. Therefore, the main objective of the present study is to identify whether it is feasible to use traditional machine learning models, including logistic regression, K-Nearest Neighbor, and random forest in stroke prediction. Among these models, the random forest outperformed other two models by classifying the sample correctly from the test dataset into binary categories. Our approach proved that simple machine learning models, especially random forest model performed exceptionally well in stroke prediction. Moreover, the classical machine learning model can also be easily implemented compared to other integrated models for stroke prediction.