Heart disease prediction using decision tree in comparison with k-nearest neighbor to improve accuracy

Abstract

The target of the task is to foresee the coronary illness by Novel Decision Tree (DT) in examination with k-Nearest Neighbor (KNN) utilizing Cleveland dataset. Coronary Disease forecasting is performed by applying Decision Tree (N=20) and k-Nearest Neighbor (N=20) algorithms. Decision Tree algorithm uses the tree structure to make decisions. K-nearest neighbor is an easy approach to solve regression and classification problems. Cleveland heart dataset is utilized for identification and prediction. The data consists of 76 attributes however, only 14 features are selected that help in diagnosing a patient healthy or affected. Accuracy of cardiovascular risk prediction using k-NN is 68.9% using decision tree is 81.9%. There exists a statistical significant difference between DT and k-NN with 0.035(p<0.05). Decision Tree algorithm appears to perform significantly better than k-Nearest Neighbor algorithm for heart disease prediction.