Performance Evaluation Of CNN, SVM, RF, LSTM, And KNN For Real-Time ECG Signal Classification In Healthcare Applications

Abstract

Heart disease is the main cause of global mortality, which demands fast and reliable clinical equipment. The study five supervised learning models- CNN, SVM, Random Forest (RF), LSTM, and KNN -For ECG Signal Classification uses, using MIT-BIH Arrhythmia DatabaseConducts comparatively. Numerical results show that CNN receives the best accuracy (92%) and fastest estimates time (0.12S), while LSTM comes closely with 90% accuracy. More than 10 independent runs confirms the importance of statistical verification performance (paired T-Test, P <0.05). The novelty of our work lies in offering quantitative evaluation of these models on the same pipeline and a quantitative evaluation of their real -time sufficiency for clinical deployment. The data for training and testing was divided 70–30, in which 5-fold cross-validation was used for strength. Experimental findings highlight the suitability of CNN for scalable healthcare applications, while LSTM and RF temporal features offer a compelling trade for systems requiring learning or noise strength.