Heart Arrhythmia Classification Using Machine Learning Algorithms

The normal heart rate in humans is between 60 and 100 beats per minute. Any variation in this range is referred to as a heart arrhythmia in medicine. This research aims to categorize arrhythmia into 16 distinct categories using SVM, KNN, and Nave Bayes like machine learning techniques. The feature-rich UCI Arrhythmia dataset is considered, which has been reduced using the Boruta feature selection technique. The consideration of only relevant features has boosted an accuracy of 9.9% in the SVM model, 3.3% in the KNN model, and 24.2% in the Nave Bayes model. Multiclass classification is performed here while taking all of the original characteristics into account from the dataset and omitting non prominent information. The performance of three algorithms is compared using accuracy and kappa ratings. After performing feature selection, the results demonstrate comparative performance between SVM (accuracy 71.3%) and Naive Bayes (accuracy 70.4%), with KNN method having the lowest accuracy of 62.6%.