Early Heart Disease (HD) detection with Deep Learning (DL) techniques can benefit the medical treatment of patients. Nowadays, life has become stressful, and living style is not so healthy resulting in increased chances of heart-related disorders. In this scenario, if the biomarkers of patients like their age, sugar level, cholesterol, heart rate, etc. are monitored regularly, then it is easy to predict heart disease or anomaly beforehand which in turn helps to facilitate the medical treatment timely. This paper proposes a novel Deep Learning Architecture (named DLA-HDD) to predict heart disease from the biomarkers recorded from patients’ medical history. The health records are fed as input to the model, which is further processed by the dense layers to extract the hidden patterns among the features of the data. The trained deep architecture is then used to predict the heart disease for future health prospects of the patient. The Kaggle dataset for heart disease (having 1025 records) is used for experimentation and an empirical comparison of the DLA-HDD model is made with prevailing Machine Learning (ML) and Deep Learning models. After Drawing the statistical comparison, it can be concluded that the proposed deep learning architecture (i.e., DLA-HDD) is the best with performance metrics of 99.7% AUC, 97.2% Accuracy, 97.3% F-measure, 98.2% Recall and 96.4% Precision among all the candidate models.