Classification of Acute Liver Failure using Machine Learning Algorithms

Abstract

With changing lifestyles, Acute Liver Failure (ALF) has been witnessed in masses lately. In order to properly diagnose and probable arrest of the ailment, this study classifies ALF using ten standard machine learning (ML) on a publicly available dataset containing 8785 data points. The dataset was divided into two sets – DF1 (containing 90% of the data), and DF2 (containing 10% of the data). DF1 was used for training and validation using a data share of 80:20 for train:validate. DF2 was used for testing. The models were further tuned which reflected a train accuracy, and F1-Score of 99.6%, and 0.996 respectively for random forest algorithm. The tenfold cross-validation accuracy was 99.3%. The test accuracy, and F1-Score using DF2 reflected a value of 99.8%, and 0.998 respectively using LGBM classifier. To the best of our knowledge, this is the first attempt to classify acute liver failure ailment.