Stroke type classification model based on risk factors using resilient backpropagation neural networks.

Stroke is a disease among the top 10 causes of death in the world and Indonesia. Stroke is divided into two, namely ischemic stroke and hemorrhagic stroke. Stroke can happen to anyone at any age with the characteristics of blood vessels experiencing blockage or rupture resulting in a crisis of blood supply that carries oxygen to the brain. The risk factors for stroke are the same as heart disease or other blood vessel diseases such as hypertension, diabetes, and cholesterol. One way to prevent stroke is to minimize several diseases that cause clogged or ruptured blood vessels. This study aims to predict the results of the classification of stroke types based on 6 factors, namely age, gender, cholesterol levels, length of stay (LOS), history, and blood sugar levels of stroke patients. The classification method chosen is one part of machine learning, namely the resilient backpropagation neural network (RBNN) because it is in accordance with the type of data used. The results showed that the prediction results of stroke type classification using six predictors were included in the good category and reached the optimum at the use of 5 nodes in the hidden layer and the resulting error was 9.26. The results of the evaluation of classification predictions using the confusion matrix also get good results, namely with an accuracy of 80% and an F1-score of 82% with a precision of 72% and a sensitivity of 95%. This result will be more optimal if there is the handling of imbalanced data. [ABSTRACT FROM AUTHOR]