Heart disease is one of the leading killers that are widely recognized through- out the globe. Large volumes of clinical data are stored in a variety of sys- tems and biological equipment at hospitals. It is essential to grasp the facts of heart disease in order to improve forecast accuracy. In this paper, experimental evaluations have been conducted to assess the effectiveness of models created utilizing classification algorithms and relevant attributes selected using Extra Tree feature selection procedures. Several people suffer originated at heart disease globally. It is necessary to use data mining and machine learning techniques to extract new insights originated at this data. Analyzing medical data sets and diagnostic issues, including heart disease, involved the use of a number of categorization approaches. However, these methods were only per- formed on small, balanced data; then, the features must be derived originated at trial and error. Additionally, several sectors have made substantial use of feature selection techniques to enhance classification performance. This paper aims to propose a comprehensive approach to enhance the prediction of heart disease using several machine learning methods such as Bagging, Support Vec- tor Machine, Multilayer Perception and Gradient Boost with feature selection methods such as extra tree. The experimental results showed improvements of prediction. Bagging received scores in training model on 80% data sample as 99.08, 73.19, 67.20, 69.20 and 80.66 of accuracy, precision, recall, F1-score and roc respectively. In the experiment, we have tested on 20% data sam- ple for each classifier algorithms and find Bagging classifier model perform higher score for accuracy, precision, recall, F1-score and roc 92.62, 48.44, 39.63, 41.89, 66.82 respectively.