Heart disease is leading cause of death. Heart disease can be detected by various medical diagnosis mainly by Electrocardiogram (ECG). The measurement of electrical activity is called ECG. Arrhythmia is the name for an unnatural heartbeat. An arrhythmia may cause serious problem. Identifying and classifying normal and abnormal (arrhythmic) heartbeats is the primary goal of the study work. This paper predicts the heart disease at early stage using deep learning approach—convolutional neural network model. The dataset collected from PTB-XL ECG database consists of 21,837 records from 18,885 patients of 10 s length 27 columns. The CNN-proposed model has achieved accuracy of 89.05%, precision has achieved 85.85%, recall has achieved of 68.29%, and loss has 0.24% when compared to other existing model achieved 80.1% and 76.9%. The proposed model has achieved good results though it has taken more datasets.