Heart Disease Prediction Using Ensemble Model

Abstract

Heart is a vital organ of the human body. The mortality rate due to cardiovascular or heart disease is in the rising trend. From the review of existing literature, timely diagnosis of the potential heart disease is the key to preventing the risk of early death. But due to the lack of proper health infrastructure in different parts of the country (especially in the rural area), timely diagnosis or the early diagnosis of heart disease has been a problem in Nepal. In this thesis, ensemble learning is used to predict the potential heart risk in patients. First different supervised learning models such as Logistic Regression, SVM, Decision Tree, KNN, and Gaussian Naive Bayes were built out of UCI heart disease dataset. These models gave accuracy of 82.46%, 87.34%, 97.67%, 89.94% and 78.57% respectively. An ensemble model was built by combining aforementioned five supervised ML models. The voting and averaging based ensemble model gave an accuracy of 96.10% and 96.43% respectively. Upon comparison, the ensemble model gave overall better accuracy. A module capable of predicting heart disease for a patient was implemented.