Gradient Boosting for Heart Stroke Prediction: Investigating Unexpected Risk Factors

Abstract

Heart stroke prediction is a critical area in healthcare, aiming to identify individuals at risk and provide timely intervention. This research leverages machine learning algorithms, including Decision Tree, Random Forest, AdaBoost, and Gradient Boost, to predict the likelihood of stroke, with Gradient Boosting delivering the most accurate results. Our analysis uncovers intriguing and unexpected relationships between stroke risk and various factors such as heart disease, hypertension, and smoking habits. Contrary to conventional wisdom, our findings suggest that individuals with lower incidences of hypertension and heart disease exhibit increased stroke risk. Additionally, non-smokers appear to have a higher likelihood of experiencing a stroke compared to smokers. Furthermore, Body Mass Index (BMI), marital status, residence type, and work type also significantly influence stroke risk. These anomalous findings necessitate further investigation to understand the underlying causes and implications. This study highlights the importance of using advanced machine learning techniques to uncover complex patterns in health data, which can lead to more effective prevention strategies.