Detection of Cardiac Arrhythmia Using Machine Learning Approaches

Arrhythmia is a cardiovascular disease that alters the heart rate, resulting in too fast, too slow, or irregular rhythms. It is a life-threatening disease if left untreated. Traditionally, arrhythmia is diagnosed by a trained doctor, using an electrocardiogram to analyze irregular heartbeats. However, these methods are vulnerable to inadvertent misdiagnosis, especially during the early stages of the disease. In this paper, an approach for cardiac arrhythmia detection is presented, where the subjects or instances are first categorized as diseased or normal and then further graded into normal (non-diseased) or as distinct subtypes of cardiac arrhythmia. The dataset was obtained from the UCI Machine Learning Data Repository, and machine learning methods such as XGBoost, CatBoost, SVM, and Random Forest, were experimented with. Addition-ally, the mutual information-based feature selection approach, minimal redundancy maximum relevance (mRMR), is proposed to improve classification accuracy. Standard evaluation metrics such as accuracy, f1-score, precision, and recall are utilized for comparison of the obtained results. The experimental results demonstrated that accuracy of 81.48% was achieved for multi-class classification, while binary classification achieved up to 84% accuracy.