The pattern of the ECG rhythm and heart rate represent the status of the cardiac heart. If appropriately evaluated, an ECG may reveal information about numerous arrhythmia illnesses of the heart. Clinical ECG observation might take several hours and be quite tiresome. Furthermore, visual analysis cannot be depended on, and the analyst risks missing critical information. The huge variety in the morphologies of ECG waveforms is the most challenging difficulty that today's automated ECG arrhythmia analysis faces. The goal of this study is to look at ECG arrhythmia classification using a neural network and numerous characteristics. The architecture demonstrated in this study, which is built on feed forward back propagation neural network with Logistic regression based weight updation coefficient, may be trained to create ECG signals that are equivalent to real-world ECG signals. The results reveal that the weight adaptation technique improves the performance of all classification networks. The suggested approach is useful in classifying various arrhythmias.