A Bayesian Optimization-Based Framework for Building Stacked Ensemble Models in Liver Disease Prediction

The early diagnosis of liver disease is vital for saving lives, managing the condition effectively, and improving patient outcomes after treatment. Now, medical diagnosis and prognosis have made heavy use of data mining tools which has immense quantity of medical data has been analyzed using these data mining approaches. Predicting liver disease at an early stage is made improved and more accurate with the use of cutting-edge machine learning algorithms. The main purpose of this study is developed a two layer staked ensemble learning with Bayesian optimization to enhance the accuracy where the dataset was collected from the Kaggle online repository which has 30,000 patient records with 11 attributes. The machine learning algorithms namely XGBoost, Decision Tree, AdaBoost, and Random Forest are applied as base learners and Random Forest is applied as the meta-learner. The total attributes of this work has reduced eleven to five by using mutual gain infommation without compromising the accuracy which is very cost effective. Finally, the findings of this research is remarkable; the accuracy is 99.99% as well as precision, recall, and F1-score, all of which are likewise 99.99% of the hybrid algorithms.