Liver Disease Prediction Using Bayesian Optimized Classification Algorithms

Liver disease poses a significant public health concern globally, necessitating accurate predictive models for early diagnosis. This research study uses the efficacy of Bayesian optimization in enhancing the performance of classifiers for liver disease identification. Specifically, it evaluates popular machine learning techniques, including Random Forest, Support Vector Machines, Adaptive Boosting, and Extreme Gradient Boosting, to determine their effectiveness. The Pearson Correlation Feature Selection algorithm is used for choosing the best features and Bayesian optimization is employed to fine-tune the hyperparameters of these algorithms. The research makes use of a Kaggle dataset from the UCI machine learning library, which includes clinical features that are utilized to train and assess the prediction models. The findings show that optimization enhances these classification systems' performance, leading to higher prediction accuracy. RF achieved the highest accuracy of 81.06% followed by SVM (80.81%), AdaBoost (77.08%), and XGBoost (79.85%).