Defining The Best-Fit Machine Learning Classifier Prediction Model For Diagnosis of Heart Disease

Abstract

Cardio vascular disease or alternatively heart disease is the primitive cause of death all around the world. Last few decades, it was observed that maximum death cases occurred due to heart failure. The heart failure death cases are associated with many risk factors for example high blood pressure, cholesterol level, sugar level etcetera. Therefore, it is advisable that regular and early diagnosis of these factors may reduce the risk of heart failure and hence achieve prompt disease management service. A commonly used technique to process these enormous medical data is called data mining, which help the researchers in health care domain. Several machine learning algorithms are used to analyses these data and help to design the best-fit model for early detection of heart diseases. This research paper contributes various attributes related to heart mal functioning and build the best-fit model using supervised learning algorithm such as various tree (fine tree, medium tree etc), Gaussian Naïve Bayes, Coarse KNN, Medium Gaussian SVM algorithms. In this paper, we used the data set from Kaggle.com. These data set comprises with total 732 instances along with 5 attributes. All these 5 attributes are to be considered for testing purpose and also to find out the best fit model for prediction of heart disease. In this research article we also compare the various classification models based on supervised learning algorithms. Based on the performance and accuracy rate we therefore, choose ‘Medium Tree’ model as the best-fit model. Maximum accuracy is obtained for ‘Medium Tree’ model. The confusion matrix for each model are calculated and analyze.