Machine Learning-based ECG Classification using Wavelet Scattered Features

Abstract

Cardiac abnormalities are one of the leading
causes of mortality and morbidity among the population.
Changes in the morphology and rhythm of the cardiac
signals associated with cardiac abnormalities need to be
identified and classified. Advances in artificial intelligence
pave the way for precise classification. The preprocessed
ECG signal segments undergo wavelet scattering to extract
the low variance features with reduced dimensions are
rearranged and the key features are selected using
Minimum Redundancy and Maximum Relevance
(MRMR) feature selection algorithms chosen by
comparatively analyzing different feature selection
algorithms and the selected features are fed to the machine
learning models. Classification of ECG signals is
comparatively analyzed using different Machine Learning
models such as Support Vector Machine (SVM), K-Nearest
Neighbor (KNN), Decision tree, and Artificial Neural
Network (ANN) models with 10-fold cross-validation. The
performance is improved by optimizing each model by
tuning the hyperparameters. Among the twenty models,
the cubic SVM model achieves the highest accuracy of
99.84 percent.