Classification of Cardiac Arrhythmia Using Supervised Learning

Abstract

Cardiovascular disease (CVD) is a leading cause of mortality, often involving cardiac arrhythmias characterized by abnormal electrical activity. The sinoatrial (SA) node serves as the heart's primary electrical impulse source. This work focuses on developing an algorithm to detect cardiac arrhythmia using morphological features extracted from Electrocardiogram (ECG) data. While ECG is a widely used non-invasive tool for cardiovascular diagnosis, it has limitations, particularly in detecting infrequent arrhythmias. The proposed supervised machine learning model aims to classify various cardiac arrhythmias, including paroxysmal atrial fibrillation and congestive heart failure, providing healthcare professionals with a valuable tool for accurate categorization and monitoring of cardiac conditions. The research involves data preparation with WEKA-based processing and feature extraction, utilizing K-Fold cross-validation for a dataset of 1200 instances and 47 attributes. Classification of Arrhythmia, Atrial Fibrillation (mostly Paroxysmal), Congestive Heart Failure, and Normal Sinus Rhythm is performed in WEKA using Naïve Bayes, Decision Tree, and k-Nearest Neighbors. All three classifiers exhibit high overall accuracy from 94% to 95.58%. Naïve Bayes slightly outperformed the others with 95.58% accuracy followed closely by J48 Decision Tree at 95.08%.