Performance Evaluation of Classification Methods for Heart Disease Dataset

Abstract

Classification techniques are greatly deployed in several application domains for the purpose of classification, estimation and prediction. This paper evaluates the performance of three basic classifiers (i.e., k-NN, Naïve Bayesian and Decision Tree) for the highly sparse dataset (i.e., medical data), which has been acquired from online machine learning repositories. The goal of the study is to evaluate performance of three basic classifiers, since each classifier works at different learning mechanism and to predict possible treatment outcomes based on well-known patterns of the treatments. The experimental results reveal that k-NN classifier produces better prediction results as compared to other two classifiers for the considered heart disease dataset. Decision Tree classifier predicts poorly for the sparse dataset, specially, when dataset comprises of varying attribute values.