Cardiovascular diseases (CVDs) pose a significant health threat to humans, and the early discovery is heavily reliant on the prevention and reduction of CVD-related deaths. This indicates that machine learning algorithms for CVD risk factor recognition is a ballooning scope. To provide a comprehensive model for predicting cardiac disease, this paper presents an approach that combines these methods. Data collection, pre-processing, and adjustment were always handled carefully to ensure that our proposed model performed well. Here, we achieve this through the pooled dataset across several sources (Cleveland, Long Beach VA, country of Switzerland, Hungarian, and Stat log). Feature selection is performed using reduction and least absolute shrinkage and selection operator (LASSO) methods, to identify significant variables. Multiple new hybrid classifiers are developed, like Gradient Boosting Boosting Technique, AdaBoost Boosting Method, the K-Nearest Neighbors Bagging Technique, Decision Tree Bagging Technique, and Random Forest Bagging Approach (Yadav and Pal in Int J Pharmaceut Res, 2020 [7]; World Health Organization and Dostupno in ‘Cardiovascular diseases using ML Algorithms [8]). These hybrids combine bagging and/or boosting with conventional classifiers during training in a hope to produce even better prediction estimates. Model performance is measured with performance metrics, such as Negative Predictive Value, False Positive Rate, False Negative Rate, and Accuracy, The sensitivity level Error Rate, Precision, and F1 Score. After careful analysis, the results are compared after an exhaustive study, it was found our proposed model under relief selecting features and the Random Forest Bagging Technique, gave an accuracy of 99.05%. These results show how well the smart phone-based approach to risk variables can help predict risk variables for heart disease leading to earlier detection and treatment.