Mortality Prediction Using Data Mining Classification Techniques in Patients With Hemorrhagic Stroke

Stroke is a major health problem in Indonesia and the world, and the cause of disability and death. Hemorrhagic stroke has a higher mortality rate compared to ischemic stroke. The objective of this study is to create a mortality prediction model by using a J48, Jrip, and Multilayer Perceptron algorithms based on the demographics data and the clinical data in patients with hemorrhagic stroke. Mortality prediction of stroke patients helps the physician to determine prognosis, targeted treatments, and prepare patients and families. 1329 subjects obtained from Hospital Stroke Registry in Yogyakarta after exclude ischemic stroke patients. This study uses 10-fold cross-validation and confusion matrix to construct and evaluate a model. The performance of J48 and Jrip is higher than Multilayer Perceptron. The use of data mining algorithms able to predict the mortality of patients with hemorrhagic stroke.