Multisurface Proximal SVM Based Decision Trees For Heart Disease Classification

This paper presents a Multisurface proximal SVM (MPSVM) based decision tree system, combined with ensemble methods of Random Forest and Gradient Boosting, for heart disease classification. Decision trees are very popular for classification tasks and normally use a set of axis-parallel decision boundaries to classify the data. The MPSVM based trees, used here, can learn decision boundaries of any orientation, and are, hence, more flexible in learning the data. The dataset used in this study is the Cleveland heart disease dataset, which is a 5-class classification problem, and has data of 303 patients, each with 13 diagnostic features. Our method is able to achieve state of the art results on the 5-class classification problem with a testing accuracy of 73.33% and on the 2-class classification problem with a testing accuracy of 91.53%.