Integrated Wavelet Decomposition and Multi-Model Approach with Fractional-Order Sparrow Search Optimization for ECG Classification

The classification of electrocardiogram (ECG) signals plays a crucial role in medical diagnostics, as accurate classification can significantly enhance the efficiency of heart disease diagnosis. This research introduces a novel method, Integrated Wavelet Decomposition and Multi-Model Approach with Fractional-Order Sparrow Search Optimization (WMFO). The method employs wavelet decomposition (WD) to decompose ECG signals into six distinct frequency components, with each component being trained using different models. The hyperparameters of these models are fine-tuned through the Fractional-Order Sparrow Search Algorithm (FOSSA), further improving the adaptability and accuracy of the models. The WMFO achieved a balanced accuracy of 93.08%, a geometric mean of 82.87%, and a Youden's index of 80.27% on the MIT-BIH arrhythmia dataset. Additionally, ablation and comparative experiments demonstrated the superior performance of this method in ECG classification tasks.