Journal Article

An effective approach for early liver disease prediction and sensitivity analysis

Hasan, Mohammad Kamrul, Ahammad, Ishtiaq, Prity, Farida Siddiqi, Uddin, Main, Prosad, Ratul, Zayed-Us-Salehin, Fatema, Sharmin, Khan, Md. Ashikur Rahman, Afrin, Faria

2023

Overview

The liver is one of the most vital organs of the human body. Even when partially injured, it functions normally. Therefore, detecting liver diseases at the early stages is challenging. Early detection of liver problems can improve patient survival rates. This research enlightens on several Artificial Intelligence techniques, including the Bagged Tree, Support Vector Machine, K-Nearest Neighbor, and Fine Tree classifier, to predict the presence of liver disease in a patient at an early stage. This study compares those models and selects the best technique to detect liver disease at an early stage. The classification performance is measured using the confusion matrix, True Positive Rate (TPR), False Positive Rate (FPR), ROC curve, and accuracy. The result shows that the Bagged Tree classifier achieves the highest classification accuracy (81.30%), which is very promising compared to the other algorithms. The proposed system also performs sensitivity analysis on the dataset to investigate the impact of each attribute on the model’s performance. It has been demonstrated that Alanine Aminotransferase (sgpt) attribute has the most significant impact on the prediction of liver disease. The proposed method could be used as an assistant framework for liver disease detection at an early stage.

• Publisher: Springer International Publishing,Springer Nature B.V
• Subject: Accuracy / Alanine / Artificial Intelligence / Classification / Computer Science / Computer Systems Organization and Communication Networks / Data Structures and Information Theory / Liver / Mathematics of Computing / Sensitivity analysis / Software Engineering/Programming and Operating Systems / Support vector machines / Theory of Computation