Feature Selection Using Binary Grey Wolf Optimization for Survival Prediction of Hepatocellular Carcinoma

The World Health Organization predicts that liver cancer will become a significant health issue worldwide and is expected to hit more than a million people by 2025. Hepatocellular carcinoma (HCC) is the primary stage of liver cancer. Survival prediction of HCC remains challenging because of complicated datasets, insufficient data, and difficulty in identifying pertinent features. This study examines the survival prediction of an HCC by proposing a methodology that combines robust imputation and feature selection methods. Missing values are handled using the k-Nearest Neighbors, Binary Grey Wolf Optimization (BGWO) is employed to identify the most impactful features for survival prediction. The preprocessed dataset is fed into the classification models such as Random Forest, Decision Tree, Naïve Bayes and k-Nearest Neighbors to predict the survival outcomes of HCC patients. The Random Forest classifier exhibited higher performance, achieving an accuracy of 95.12% using 14 selected features. The study evaluates the models based on standard metrics and their applicability to unseen data. The proposed framework provides an efficient approach to improving survival predictions for HCC patients, contributing to better disease management and treatment planning.