Development of Hepatitis Disease Detection System by Exploiting Sparsity in Linear Support Vector Machine to Improve Strength of AdaBoost Ensemble Model

Abstract

Hepatitis disease is a deadliest disease. The management and diagnosis of hepatitis disease is expensive and requires high level of human expertise which poses challenges for the health care system in underdeveloped and developing countries. Hence, development of automated methods for accurate prediction of hepatitis disease is inevitable. In this paper, we develop a diagnostic system which hybridizes a linear support vector machine (SVM) model with adaptive boosting (AdaBoost) model. We exploit sparsity in linear SVM that is caused by <a:math xmlns:a="http://www.w3.org/1998/Math/MathML" id="M1"> <a:msub> <a:mrow> <a:mi>L</a:mi> </a:mrow> <a:mrow> <a:mn>1</a:mn> </a:mrow> </a:msub> </a:math> regularization. The sparse <c:math xmlns:c="http://www.w3.org/1998/Math/MathML" id="M2"> <c:msub> <c:mrow> <c:mi>L</c:mi> </c:mrow> <c:mrow> <c:mn>1</c:mn> </c:mrow> </c:msub> </c:math> -regularized SVM is capable of eliminating redundant or irrelevant features from feature space. After filtering features through the sparse linear SVM, the output of the SVM is applied to the AdaBoost ensemble model which is used for classification purposes. Two types of numerical experiments are performed on the clinical features of hepatitis disease collected from UCI machine learning repository. In the first experiment, only conventional AdaBoost model is used, while in the second experiment, a feature vector is applied to the sparse linear SVM before its application to the AdaBoost model. Simulation results demonstrate that the strength of a conventional AdaBoost model is enhanced by 6.39% by the proposed method, and its time complexity is also reduced. In addition, the proposed method shows better performance than many previously developed methods for hepatitis disease prediction.

Topics & Concepts