Prediction and Classification of Cardiac Arrhythmia Using a Machine Learning Approach

Abstract

Arrhythmia is a very common cardiac problem that occurs frequently all over the world. In order to diagnose early stage arrhythmia, this paper explores nine different machine learning approaches—Logistic regression, latent Dirichlet allocation, quadratic discriminant analysis, k-nearest neighbours, support vector classifier, support vector machine, decision tree and extreme gradient boosting (XGBoost) and proposes a model to predict patient’s chances of getting arrhythmia condition. Arrhythmia can be predicted and classified by using data from the electrocardiogram (ECG) of the patient. The algorithms classify the acquired data in 1 of 5 classes of cardiovascular Arrhythmias, where UCI Arrhythmia dataset is considered for training and testing. Results are studied and compared.