Background <p>The signals on the electrocardiogram (ECG) are crucial in diagnosing cardiac arrhythmias, which are irregular heart rhythms that indicate acute and chronic cardiovascular diseases. Proper identification of arrhythmias is vital for efficient clinical management, but this process is often time-consuming and prone to human error. Due to the rapid progress in machine learning and signal processing, automated ECG-based diagnostic systems have become reliable devices in cardiac health monitoring.</p> Methods <p>The paper presents a deep learning-based model for classifying cardiac arrhythmias. During pre-processing, the Discrete Wavelet Transform (DWT) is used to clean and decompose ECG signals into their significant frequency components. R-peaks are detected and their width is determined to identify the QRS complex, a considerable area in the recognition of arrhythmias. The Artificial Bee Colony (ABC) algorithm is then used to optimise the features to identify the most significant attributes in the QRS complex. These optimized features are then input into a Deep Neural Network (DNN) model, which is trained to classify multi-class arrhythmias.</p> Results <p>The combination of the DWT-ABC-DNN model was found to be better, with an overall accuracy of 98.779%, a sensitivity of 98.67%, and a specificity of 98.87%. These results demonstrate that the proposed method outperforms traditional classification methods, underscoring the strength and accuracy of the suggested system.</p> Conclusion <p>The proposed ECG classification model is a deep learning model that offers a very accurate and efficient method of arrhythmia detection. By combining wavelet-based pre-processing, bio-inspired optimization, and deep neural network learning, the approach provides an effective and trustworthy platform for automated cardiac diagnosis, with a high likelihood of clinical implementation.</p>