Advanced data balancing techniques with machine learning models for acute liver failure prediction

Abstract

Amongst various diseases, one of the severe diseases is acute liver failure (ALF) and it is a quick decline in liver health that normally lasts a few days to a few weeks. Machine learning (ML) techniques can play a valuable role in the diagnosis and management of ALF. The proposed study made an effort to remedy the issue of the Kaggle Dataset's class imbalance by carrying out an exhaustive experimental assessment making use of two distinct approaches, namely synthetic minority oversampling technique (SMOTE) and synthetic minority oversampling technique and edited nearest neighbours (SMOTE-ENN). Both SMOTE-balanced and SMOTE-ENN balanced datasets are used to train the support vector machine (SVM), K-nearest neighbors (KNN), logistic regression (LR), decision tree (DT), random forest (RF), eXtreme gradient boosting (XGBoost), and stacking models. Compared to the SMOTE method, the results demonstrated that the SMOTE-ENN balanced dataset achieved a considerable increase in the accuracy of its predictions. The results showed that the KNN algorithm has attained 99.52\% accuracy, along with a precision of 99.07\%, recall of 99.35\%, and F1 measure of 99.04\%. As a result, we discovered that a data balancing method that is not overly complicated and a supervised ML algorithm could be used to forecast ALF with very high accuracy and excellent potential for utility.