Preprocessing Unbalanced Data using Support Vector Machine with Method K-Nearest Neighbors for Cerebral Infarction Classification

Abstract

Cerebral infarction is focal brain necrosis due to complete and prolonged ischemia that affects all tissue elements, neurons, glia, and vessels. Stroke infarction or known as cerebral infarction is a condition of damage in the brain due to insufficient oxygen supply, due to obstruction of blood flow to the area. Research shows stroke infarction does not only occur in the elderly, but occurs at a young age of around 15-55 years, especially with certain risk factors, such as diabetes, hypertension, heart disease, smoking, and long-term alcohol consumption. In diagnosing the presence of cerebral infarction in the brain, machine learning is used because it is not enough just to use a CT scan to diagnose. Therefore, it requires timely detection and more accurate methods of classification. This study aims to use Support Vector Machine (SVM) as preprocessing and K-Nearest Neighbors (KNN) algorithm to classify Infarction Cerebral. In this study, discusses the application of SVM to deal with class imbalances. The first strategy is to balance data using SVM as a preprocessor and the actual target value of the training data is then replaced by trained SVM predictions. Then, the modified training data is used to classify with K-NN method. We use data CT scan result from a Department of Radiology at Dr. Cipto Mangunkusumo Hospital (RSCM). This accuracy in this paper shows around 69,85 %.