ECG classification with learning ensemble based on symbolic discretization

This paper introduces a novel learning ensemble algorithm designed for the classification of Electro-Cardio Graphic (ECG) signals. In real-time monitoring of cardiovascular patients, addressing the scalability challenge requires an adapted representation that enhances dimensionality reduction before the classification process. Our approach focuses on a discretization technique that transforms Time Series (TS) data into a sequence of ordered symbols, thereby enabling simultaneous compression and classification of ECG signals. Experimental results conducted on various ECG databases from the UCR archive benchmark demonstrate a significant improvement over two types of classifiers, namely distance-based and structure-based, and competitive results when compared to shapelet-based classifiers. The proposed algorithm and technique hold promise for enhancing the efficiency and accuracy of ECG signal classification, which is vital for the timely diagnosis and treatment of cardiovascular diseases.