Predictive Modelling of Stroke Occurrence among Patients using Machine Learning

Abstract

Stroke is a global public health concern with severe consequences. Early detection and accurate prediction of stroke occurrence are crucial for effective prevention and targeted interventions. This study proposes a machine learning-based approach to predict the likelihood of stroke among patients. A comprehensive dataset encompassing demographic, clinical, and lifestyle factors of a large patient cohort was employed. Variables such as age, gender, hypertension, diabetes, smoking status, BMI, and medical history were considered. Advanced machine learning algorithms, including logistic regression, decision trees, random forests, and support vector machines, were utilized to analyses the dataset and develop a predictive model. The results demonstrate that the machine learning-based approach achieved high predictive accuracy in identifying individuals at risk of stroke. The model exhibited excellent sensitivity and specificity, enabling effective stratification of patients based on their stroke likelihood. Developing an accurate stroke prediction model using machine learning holds immense potential for proactive healthcare strategies and personalized patient care. Early identification of high-risk patients enables timely intervention and implementation of preventive measures, potentially reducing the burden of stroke-related complications. This study showed that the supervised K-Nearest Neighbors Algorithm (K-NN) model outperforms the other methods, with an accuracy of 95% compared with other models.