AUTOMATING THE DIAGNOSIS OF LIVER DISEASE USING MACHINE LEARNING

Abstract

Liver diseases are of significant health concern to the world as they are usually fatal because of delayed diagnosis and ineffective screening. To enhance clinical outcomes, proper and prompt diagnosis is thus important. This paper proposes an automated machine learning-based system of liver disease diagnosis, based on systematic data preprocessing and balancing of classes. To assess a number of machine learning models i.e., LightGBM (LGBM) and Decision Tree (DT), the Indian Liver Patient Dataset (ILPD) of the UCI Machine Learning Repository was utilized. The data was prepared with duplicate removal, filling of missing data using Multivariate Imputation by Chained Equations (MICE), Z-score normalization for the detection and removal of the outliers. The problem of class imbalance was solved using Synthetic Minority Oversampling Technique (SMOTE). According to experimental findings, LGBM model performed best, having accuracy 74.77%, precision 78.72%, recall 67.89% and an F1-score of 72.91%. The results indicate that outliers removal and balanced representation of data is effective in training the machine learning models and subsequent effective prediction of liver disease, as well as in the construction of intelligent decision-support systems in clinical practice.