In healthcare field, cardiovascular disease prediction is the challenging task. It is a challenging task to predict cardiac disease with high accuracy. Recently, a number of ML-based systems for the forecast and analysis of heart illness have been introduced. However, systems can’t manage large amount of datasets. To overcome above issues, a stacked ensemble net is proposed to provide highly accurate outcomes and enhances the efficiency of the performance without any high-dimensional dataset. Using a deep learning model, heart disease can be easily predicted, and doctors can diagnose heart disease and provide better treatment to the patients. The synthetic minority over-sampling technique and Edited nearest neighbor (SMOTE-ENN) accomplish oversampling and downsampling at the same time, and it balances the imbalanced dataset. The features are extracted from stacked DBN and stacked ANN, which is given to ensemble techniques such as decision tree (DT), bagging and K-Nearest Neighbors (K-NN). The recommended model is called a stacked ensemble net, which combines the stacked deep belief network (DBN) and stacked artificial neural network (ANN). It eliminates the possibility of data leaking and over fitting. The present work is estimated with cardio illness data and related to traditional classifiers on the basis of feature extraction and ensemble methods. Predictive methods such as DT, Bagging, K-NN, Adaboost, Gaussian naïve Bayes, Random forest, SVM and Logistic regression are estimated according to accuracy, recall, precision, and AUC. The proposed system acquires an accuracy of 95.8%, which is greater than state-of-art techniques. This study displays that, in contrast to other state-of-art techniques, the proposed work is more effective at predicting heart disease. Keywords: Heart disease, stacked deep belief network (DBN), stacked artificial neural network (ANN), stacked ensemble net, decision tree (DT), bagging, k-nearest neighbour (K-NN).