How to increase the effectiveness of the hepatitis diagnostics by means of appropriate machine learning methods

Abstract

This paper presents how to improve the diagnostic process of hepatitis B and C based on collected questionnaires from patients hospitalized in all regional departments of infectology in Slovakia. Performed experiments were oriented in two directions: economic demands of the recommended treatment based on realized diagnostics and possible improvement of hepatitis diagnostics by means of exploratory and predictive analysis of additional information provided by patients. Exploratory data analysis was used to confirm or to reject some expected relationships between input attributes (e.g. ager or gender) and target diagnosis. Also, predictive mining resulted into interesting decision rules that can be used in medical practice as supporting information at an early stage of the diagnostic process. Finally, analysis of the treatment economic demands based on the estimated costs showed the need for timely and quality diagnostics to minimize the percentage of patients for which was hepatitis diagnosed late.