Relief Feature Selection and Bayesian Network Model for Hepatitis Diagnosis

Abstract

A doctor diagnose a disease by evaluating patient condition or by comparing with another patient that have similar conditions or symptoms. In computer science, this task can be done by a computer program that included intelligent algorithm in it. Some disease have similar symptoms, such as typhoid fever, hepatitis, and dengue fever. Based on UCI database there are 17 symptoms of Hepatitis that may be similar with other disease, so it needs a method to find the major symptoms. In this research, we proposed hepatitis diagnose using statistic Bayesian network and find major symptoms using ReliefF algorithm. ReliefF algorithm resulting 4 majority symptoms and used to constructing Bayesian Network. ReliefF and Bayesian Network have 76,8% accuracy, 76,5% precision, and 100% recall for 69 test data.

Keywords: Hepatitis, ReliefF, Bayesian Network, Probabilistic.