PCO: A Novel Parrot–Cheetah Optimization Framework for Automated Arrhythmia Detection from ECG Signals

Abstract

Detecting cardiac arrhythmias early and correctly from an electrocardiogram (ECG) can prevent serious, sometimes fatal, cardiac events. Doing this automatically is hard: ECG recordings are noisy, the classes are usually imbalanced, and the feature set extracted from the signal is large. This paper presents a hybrid optimizer, the Parrot–Cheetah Optimizer (PCO), for arrhythmia detection. PCO pairs the broad search of the Parrot Optimizer (PO) with the fast local search of the Cheetah Optimizer (CO) and switches between them whenever the search stops improving. Following preprocessing, we extract time-domain, frequency-domain, and nonlinear features, and PCO then selects a small subset of these features before a supervised classifier labels each beat. On a standard ECG dataset, PCO outperformed the metaheuristic optimizers considered here in accuracy, sensitivity, specificity, and F1-score, and came close to recent deep learning classifiers while using far fewer features and less computation. Its results also varied little across repeated runs. Together these properties make PCO a practical choice for real-time arrhythmia monitoring.