Effect of Feature Selection by Genetic Algorithm on Early Prediction Performance of PAF Attack

The heart is very important to pump in a healthy way, but any disease that can occur in the heart has vital preventive measures. One of the most important of these diseases is Atrial Fibrillation (AF). This disease is a disturbance caused by excitations that occur outside of the sinoatrial node that occurs in the atrium of the heart. Paroxysmal Atrial Fibrillation (PAF) is the first stage of AF. Early prediction of this disease prevents the disease from passing to the other heavier stages. In this study, it was aimed to develop a warning system that warns PAF patients before an attack begins. Starting from the PAF, 99 pieces of data consisting of 10 parts in 5 minutes were used. Time domain measurements and poincare plot measurements were obtained over the data. the features that best distinguish the classes have been determined by choosing a feature with a genetic algorithm. As a result, PAF can be predicted up to 7.5 minutes before the attack occurs using the selected features.