A New Expert Hepatitis Diagnosis System Based on Linear Discriminant Analysis-Extreme Learning Machine Classifier

Abstract

In this study, a new expert hepatitis diagnostic system based on the Linear Discriminant Analysis (LDA)-Extreme Learning Machine (ELM) Classifier method is suggested. ELM has various applications in fields such as biomedical engineering, computer vision, system identification and robotics. In this paper, a detailed explanation of the ELM development is provided. The UCI machine learning database was used for the hepatitis illness dataset. Proposed method performance is evaluated through statistical methods such as classification accuracy, sensitivity and specificity statistical analysis methods. The structure of this hepatitis diagnosis system can be described in three phases. In first phase, the hepatitis dataset is obtained and features reduced. These 19 features of the hepatitis dataset are reduced to 10 features using Linear Discriminant Analysis (LDA), Principle Component Analysis (PCA) and Generalize Discriminant Analysis (GDA) methods respectively. In the second phase or classification stage, each analysis reduced feature set is given to the Extreme Learning Machine (ELM) classifier. This phase indicated that the Linear Discriminant Analysis (LDA)-Extreme Learning Machine (ELM) Classifier method outperformed the two other methods, PCA-ELM and GDA-ELM. Finally, in third phase, the diagnosis performance of our LDA-ELM expert system for diagnosis of hepatitis is calculated. The resulting classification accuracy of this system was 95.17 %. In this study, results are compared to hepatitis diagnostic approach studies using the same or like dataset and conclude that the high classification accuracies. © 2019 IEEE.