Cardio-vascular disease prediction based on ensemble technique enhanced using extra tree classifier for feature selection

Abstract

Cardio vascular disease is the major death factor in the last decade. Most of the patients diagnosed with the CVD at the later stage die even though advanced treatments are given. Earlier identification of heart disease may reduce the death rate. The cost of medical diagnosis makes it perverse for the large number of people to the early fix of the disease. Machine learning and Data mining techniques are successful in medical diagnosis through non-invasive methods. In developing such models, Feature selection is very important since it affects the accuracy of the diagnosis. In this research, the feature selection is done through Extra Tree classifier method for identifying the most important feature combination for predicting the heart disease. Cleveland and Statlog datasets are used for developing and testing the model. Base classifiers such as Support Vector Machine (SVM), K Nearest Neighbour (KNN), Decision Tree, Logistic Regression, Naïve Bayes and Vote are tested over all original 13 features from datasets, 9 feature combination and 6 feature combination. It is observed that Vote classifier using 9 and 6 feature combination gives the best accuracy and F1 score.