Analysis of Stroke Classification Using Random Forest Method

Abstract

Stroke is a disease in which the sufferer experiences or experiences a rupture of a blood vessel in the brain so that the brain does not get a blood supply that provides oxygen. Patients who suffer from stroke will experience cognitive disorders ranging from decreased consciousness, visuospatial disorders, non-verbal learning disorders, communication disorders, and reduced levels of patient attention. Data from the World Stroke Organization shows that there are 13.7 million new stroke cases every year, and about 5.5 million deaths occur due to stroke. This research aims to analyze the attributes of any variables that affect the classification of strike disease and to test the performance of stroke classification in the form of accuracy, precision, recall, and f-measure. The method used is a random forest using a tree, namely 50, 100, 200, and 500. The classification of stroke is divided into stroke and no stroke. The data used is 5110, divided into 70% training data and 30% testing data. The results showed that the performance of a random forest using 100 trees was better than using 50, 200, and 500 trees, with an accuracy value of 86.82%, a precision of 15.76%, a recall of 38.15%, and an f1-score 22.30% after doing SMOTE.