THE IMPACT OF FEATURE SELECTION ON THE PROBABILISTIC MODEL ON ARRHYTHMIA DIAGNOSIS

Abstract

Arrhythmia is a type of cardiac illness identified by an irregular heart rhythm that can be either too rapid or too slow. An electrocardiograph method is required to diagnose arrhythmia. Electrocardiogram, ECG, is the result of this Electrocardiograph process. The ECG is then utilized as a diagnostic tool for arrhythmia. Because the ECG data is so extensive, an adequate processing procedure is required. Understanding the ECG data can be done in various ways, one of which is classification. Naïve Bayes is a classification technique that can handle enormous amounts of data. ECG data has a lot of characteristics, which makes classification more difficult. Feature selection can be used to eliminate non-essential features from a dataset. This research aimed to determine the feature selection’s impact on the Naïve Bayes classification. It was proven by increased accuracy by 4%, precision by 0.13, recall by 0.13, and f-measure by 0.14. The computation time was 0.03 seconds faster. The highest performance was obtained by classification with 80 features. The accuracy was 93%, precision and recall were 0.45, f-measure was 0.42, and computation time was 0.10 seconds.