Classification of ECG Signal by using Machine Learning Methods

In this study, an application of Artificial Neural Networks (ANN), Support Vector Machines (SVM), and k-Nearest Neighbor (k-NN) machine learning methods is performed to measure the classification performance of the models on classifying electrocardiogram (ECG) signals as normal and abnormal. In this scope, ECG records were obtained from an open-accessible database (PTBDB). A feature set was generated by extraction the morphological and statistical features of 80 normal and 442 abnormal ECG recordings obtained from the database, first. The feature set was applied as the input to ANN, SVM, and k-NN classifiers. The 10-fold cross-validation method was employed in the experiment in order to achieve more generalized results. As a result of the experimental study, the best classification performance was achieved using SVM, and 85.1% of accuracy, 89 of sensitivity and 51,7 specificity values were obtained. SVM was superior to other classifiers.