Arrhythmia, an irregularity in the rhythm of heart, is a life-threatening cardiac fault which causes death. The cardiovascular diseases (CVD) high-risk patients are provided with computerized Electrocardiogram (ECG) devices so as to detect Arrhythmia and to save life. Several methods were presented since last decades for automated detection and classification of ECG beat. Due to the difficulty of weak ECG signal, poor anti-interference ability, and presence of noise, diagnosing arrhythmia is challenging. So as to overcome the issues, a deep learning-based recognition and classification model is proposed. The main aim of the proposed method is to present an effective automated scheme for the recognition and classification of Arrhythmia from ECG signal. At first, input dataset MIT-BIH is considered, and preprocessing is carried to remove the noise which involves DC drift, normalization, and transfer function of LPF. After that, QRS detection is made so as to identify the peaks of filtered ECG signal. The process of segmentation is carried by means of Markov random field (MRF) Ayed model. By this MRF computation, the features are extracted. To conclude, the meta-heuristic AlexNet classifier (MH-AlexNet) thus recognizes and classifies cardiac arrhythmia diseases, thereby predicting the heart disorder type. The proposed method is validated by evaluating performance in terms of accuracy, precision, recall, and F1-score. The attained outcomes reveal that the proposed model is effective than the existing strategies by offering enhanced results.