The prediction and feature importance analysis of stroke based on the machine learning algorithm

Abstract

The prediction of the probability of the patients stroke is a challenging task in the past decades. This study aims to predict the probability of stroke in patients using machine learning algorithms. Logistic regression model was used in this study to build the prediction model. In addition, the data preprocessing technology e.g. missing value processing and feature encoding was also carried out. The dataset collected from the Kaggle Platform used for the analysis contains various clinical and demographic features of the patients. The model achieved an accuracy of 96.3% in predicting stroke probability. Furthermore, the feature importance analysis was conducted to identify the most significant features that contribute to the prediction. The results demonstrated that some features such as age, glucose level, work type and hypertension were the most important features for predicting stroke probability. The findings of this study could help healthcare professionals in identifying high-risk patients and providing timely interventions to prevent stroke occurrence.