Cardiac arrhythmia is one of the reasons of cardiovascular disease. Early and correct detection of arrhythmias, characterized by irregular and rapid heart rates, can notably reduce the risk of strokes. Analyzing an electrocardiogram (ECG) is the most frequently method used for detecting cardiac arrhythmias, because of its non-invasive nature. Manual process for arrhythmia detection is time consuming and error driven due to large and massive amounts of ECG data; computer-aided system by applying deep learning model provides early and automated identification of cardiac arrhythmia which leads better treatment toward patient. This paper presents a Convolutional Neural Network (CNN)-based deep learning model for arrhythmia classification, along with a comparative evaluation of Support Vector Machine (SVM) and AdaBoost algorithms. The experiments were conducted using the MIT-BIH Arrhythmia Database and the St. Petersburg INCART dataset. Classification of arrhythmia into the five classes mentioned as: N non-ectopic, S supraventricular ectopic, V ventricular ectopic, F fusion, and unknown Q beats. Accuracy of CNN model on MIT-BIH dataset 99.28%, SVM classifier is 95.59% and AdaBoost classifier is 89.42. The CNN model exhibited superior performance due to its ability to capture spatial hierarchies in the ECG signals. Performance of these classifiers is also compared with St. Petersburg dataset for classifying the arrhythmia along with performance measure factors precision, accuracy, recall, and F1-score.