Stroke detection using machine learning methods

Stroke is defined as the sudden interruption of blood flow to the brain. The cells in the area where blood flow is cut off can gradually lose their functionality, leading to permanent damage in the affected area of the body. Stroke can be influenced by various factors, such as age, occupational status, certain chronic diseases, and the presence of individuals in the family who have previously suffered from it. Because the assessment of these factors and the estimation of stroke risk can be quite costly and time-consuming, the likelihood of an individual facing permanent damage increases. In today's technology, the latest advancements in Artificial Intelligence and Machine Learning models allow for decisions regarding the presence of risk to be made in seconds by improving on millions of data points. This study aims to save time and protect human health from this danger by using Machine Learning methods such as Logistic Regression, K-Nearest Neighbors, Support Vector Machines, and Decision Trees to reliably determine whether an individual is at risk of having a stroke. As a result of the study, it was found that the method achieving the highest accuracy on the dataset was the Decision Trees method, with an accuracy of 91%. The ranking of the methods based on accuracy is as follows: Support Vector Machines at 89%, K-Nearest Neighbors at 81%, and finally, the Logistic Regression method at 75%.