A MACHINE LEARNING-DATA DRIVEN ANALYSIS FOR NON ALCOHOLIC FATTY LIVER DISEASE

Abstract

Nonalcoholic fatty liver disease (NAFLD), and its progressiveness of disease, such as nonalcoholic steatohepatitis (NASH) and fibrosis, show worldwide challenge, solely due to the difficulty in achieving early diagnosis given the limited initial symptoms. This research explores the data-driven machine learning (ML) to ,enhance the accurateness of classification, prediction in fatty liver disease.[1] Workflow frequently uses datasets such as the Indian Liver Patient Dataset (ILPD), electronic health records (EHR), and multi-omics cohorts (transcriptomics and lipid omics)[2]. Data preprocessing is a initial step, often involving multivariate imputation or iterative imputation for missing values, applying transformations to correct skewedness,[3] .To optimize the model efficiency and performance, feature extraction and selection methods such as the spearmen coefficient correlation was applied not linear data and Pearson coefficient to liner data.[4] Stacking Ensemble Learners and Gradient Boosting algorithms (XGBoost), consistently exhibit higher performance in prediction tasks. But for primary level 6 basic algorithms were used for prognostics.[5], [6] Detection of stages of fibrosis and differentiation between NASH and NAFL was to find disease in early stage without invasive process. Key clinical factors identified as having significant predictive value for progression 3

include prior mean and maximum AST/ALT ratio , FIB4, and metabolic and liver marker levels, and prior mean and standard deviation[7] . Functional enrichment analysis also emphasized the significant accumulation of fat in liver, scarring and then converting it into cancer[8] . Collectively, these data-driven ML approaches offer efficient means to improve early detection, refine risk stratification, and elucidate the underlying pathogenesis of fatty liver disease.