Analysis of Arrhythmia from Electrocardiogram (ECG) Data Using ML Framework

Arrhythmia is one of the life-threatening condition where the heart rhythm will either be fast or slow which is known to be as arrhythmia. In order to detect arrhythmia, there are several methods but most widely used method is Electrocardiogram (ECG) where ECG will give information about electrical signals of heart by which doctors can perform action upon. Manually analyzing the ECG recordings by doctors requires lots of time, so it is important to build a model which will evaluate the ECG signals in short period of time and with higher accuracy. Addressing this issue this paper presents a machine learning methods and algorithms such as logistic regression, decision tree, random forest, k nearest neighbor, naive Bayes, support vector machine will be applied in the collected data. The UIC dataset was used, and 9 copies of UIC dataset were made where feature values were changed randomly to chuck weather the accuracy will get increased or not. As a result algorithms will be compared and best performing algorithm for the given dataset with high accuracy and less loss will be obtained. In the obtained result the average accuracy of machine algorithms is RF (97.14%), DT (75.78%), LR (70.29)% and SVM(70.76%), KNN (73.7%), NB (72.52%). Out of 10 datasets D7 had good accuracy D4 had least.