Exploring Stroke Risk Identification by Machine Learning: A Systematic Review

Abstract

This work aims to systematize previous studies on stroke risk identification and its relationship with machine learning. A systematic review was conducted using the Web of Science and Scopus databases. The information was organized into three sections: stroke risk factors, data preprocessing techniques and techniques for identifying stroke risk with an emphasis on the most important features. The main results are as follows: risk factors are divided into modifiable (work environment and air pollution) and non-modifiable (sex, family history). The most commonly used data preprocessing techniques are SMOTE, standardization and value elimination/imputation. The most commonly used techniques for identifying stroke risk include support vector machine, random forest, logistic regression, naïve Bayes, k-nearest neighbors and decision tree.