Abstract
The topic predictive analytics is the ray that lightning the way to patch the gap between accuracy in decision-making by the expertise and the inexperience. In particular, the health domain is more crucial about disease prediction accuracy. The disease diagnosis by clinical practitioner correlates to his exposer toward the clinical observations of the disease. However, the perceptions of an experienced clinical practitioner on a medical record often fail to identify the premature states of the disease, which costs patient life in the sector of critical diseases such as heart diseases. Hence, contemporary computer science engineering research has more attention to define substantial predictive analytics built by machine learning toward heart disease prediction. The critical objective to define predictive analytics with minimal false alarming is centric to potential training data corpus, and the optimal feature selection. In order to these arguments, the contribution of this manuscript aimed to portray the feature selection approach to perform supervised learning and label the given patient record is prone to heart disease or not with minimal false alarming. The contribution is a dynamicn-gram Features Optimization by Discrete Weights of the feature correlation. The experimental study signified the performance of the proposed model compared to the contemporary methods of feature selection for heart disease prediction.