A Classification of ECG Arrhythmia Analysis Based on Performance Factors Using Machine Learning Approach

Electrocardiogram is a diagnostic tool that makes a record of the muscular and electrical activities of the heart for showing that data as a trace on a piece of paper, which is then carefully studied and interpreted by a clinical assistant. Classification of ECG signals on performance factors like sensitivity, specificity and accuracy using machine learning techniques can provide substantial input to doctors for essential and effective treatment of the patient. Arrhythmia refers to the condition when the heart beats improperly. Irregular, fast or slow beating of heart comes under arrhythmia. It is caused due to improper working of electrical impulses of the heart. In this research work, various machine learning techniques were made use for the purpose of classification of normal and arrhythmic beats. The major objective of this work is automated classification of normal and irregular beats. Introduced approach deals with a good volume of standard ECG time-series data as inputs to various machine learning classifiers such as Naive Bayes, support vector machine, ada-boost, random forest, decision tree and k-nearest neighbor classifiers. MIT-BIH arrhythmia database has been made use for the purpose of examining the best classification performance, and one of the most popular datasets of the same database has been taken for training and testing purposes. This comprises of an exact 47 records of thirty minutes each and forty percent of the entire records were those of cardiac patients. Three major performance evaluation parameters have been examined by the researchers on distinct machine learning algorithms. They are accuracy, sensitivity and specificity. A confusion matrix has been made use for describing all these measurements as false positive [FP], false negative [FN], true positive [TP] and true negative [TN]. The efficiency and potential of different methods of arrhythmia detection are demonstrated, and numerous comparisons with various other conventional machine learning techniques are also made. The researchers calculated the sensitivity, specificity and accuracy and found the best result using decision tree classifier with 88.2% accuracy.