Hybrid Deep Learning and Machine Learning Approach for ECG-based Arrhythmia Classification and Multi-Class Disease Identification

Abstract

This Study shows a thorough hybrid technique for the classification of arrhythmias utilizing ECG signals, incorporating both Deep Learning (DL) and Machine Learning (ML) methods. We assess a pasture of ML classifiers to decide the most efficient model for arrhythmia discovery, with Adaboost discovered as the dominant classifier, attaining a precision of 96.9%. In the deep learning section, we exploit VGG16 for robotic feature extraction, succeeded by classification of these features applying a Support Vector Machine (SVM), which outcome in an accuracy of 96.6%. To develop an expansion of our study, we concoct a multi-class disease classification approach predicate on a hybrid CNN-LSTM framework, which attains an impressive accuracy of 99.53%. These experiment outcomes feature the conclusiveness and strength of our hybrid approach in enhancing ECG-based arrhythmia identification and multi- class CVD classification..