An automated detection of heart arrhythmias using machine learning technique: SVM

Abstract Electrocardiogram (ECG) is widely used technique in study of heart beat irregularities such as cardiac arrhythmias, sinus rhythms and heart failure. It is a significant and popular technique to classify and detect the cardiac infraction. ECG signal analyses the electric activity of heart and outputs it in the form of waveforms which help in detection of heart irregularities. The main goal of this research work is to classify the arrhythmia with more accurate results in less computational time. The research is carried in machine learning technique- SVM classifier using Discrete Wavelet Transform (DWT). In this methodology, ECG samples of three different classes-Normal Sinus Rhythm, Congestive Heart Failure and Cardiac Arrhythmia were collected from MIT-BIH and BIDMC databanks. The collected signals were prepared into training set and testing set with a ratio of 70:30 percent respectively. Total 190 features were extracted from the prepared data using Discrete Wavelet Transform. DWT was chosen as it has the ability to vary the window size depending on the frequency. The extracted features were given to SVM classifier, which is best for classification purpose. The results were evaluated using the testing set and the final results were plotted using a confusion matrix. The performance accuracy of the model is 95.92 percent.