Cardiac arrhythmia diagnosis applying machine learning techniques

Abstract

Cardiac arrhythmia affects millions of people worldwide. Although commonly occurring, identifying and correctly diagnosing is not a simple task. In this context, this paper presents a study on the application of Machine Learning to the identification and diagnosis of cardiac arrhythmias. Classifiers were obtained using the k-NN and SVM algorithms, and tests were performed using data from the Arrhythmia dataset, which consists of information obtained from patients' ECG examinations, as well as information related to their lifestyles. Three tests were performed; in the first one, the ability of classifiers to identify whether or not an arrhythmia episode occurred. In the second one, the performance of the classifiers in the identification of the arrhythmia type was verified, and in the third one, the investigation was performed considering the gender of the individuals. The results indicate that the use of Machine Learning may, in fact, assist specialists in the diagnosis of arrhythmias. In all tests k-NN presented better performance when compared to SVM. The best result among all tests was obtained by gender classification, where k-NN presented a accuracy of 94.03% in identifying arrhythmia occurrences in female patients.