Predicting Heart Attack Using NBC, k-NN and ID3

Abstract

We are living in a world full of data. Every day people encounter large amounts of data. Main problem here is dealing with this huge data. Data mining techniques can be used to handle such huge data. Health care environment collects vast amounts of data, but the unfortunate thing is that it is not efficient in extracting useful information from this wealthy data. Data mining techniques can be applied to extract valuable knowledge from the health care environment. In this paper, three supervised learning classification algorithms have been implemented to predict heart attack risk from heart disease database. The classification algorithms used are Naive Bayesian Classification (NBC), k-Nearest Neighbor (k-NN) Classification and ID3 Classification. As a pre-processing step Discretization of continuous variables is adopted. The heart disease data set is trained with these classifiers. A GUI is designed so that the user can input patient�s record. The system is then able to predict whether or not the user has a risk of heart attack. The performance of these three algorithms is determined by computing accuracy. From the experiments, it is found that ID3 Classification outperforms other two classifiers with the accuracy of 91.72%.