Performance Evaluation of Machine Learning Approach for hepatitis Classification with genetic Algorithm feature Selection

Abstract

Hepatitis is an inflammation of the liver commonly caused by viral Hepatitis. It is a global disease that causes deaths. Machine learning methods have been successful in hepatitis disease diagnosis. Some of the machine learning methods proposed for diagnosis and prediction of hepatitis disease have demonstrated a significant ability in achieving good performance, however some machine learning algorithms do better than others with regards performance accuracy of the diagnosis of Hepatitis disease using the datasets available. In this paper, we introduce genetic algorithm for feature selection in effort to improve performance, three classifiers including Support vector machines (SVM), K-nearest neighbors, and Logistics regression. Ten out of 19 features were created using Genetic algorithm which was run for ten generations and used as the selected features for the classification of Hepatitis disease sourced from UCI data repository. The experiment was accrued out on the datasets using selected features subsets created using Genetic algorithm. The result of our analysis shows that KNN outperformed the other classifiers.