Predicting Liver Disease Using Decision Tree Ensemble Methods

Özet

Damages that may occur in the liver, which has an important task for the human body, can cause fatal consequences. For this reason, early diagnosis of liver disease is important. In this study, liver disease was tried to be diagnosed by using Ensemble learning methods, depending on several clinical values obtained from liver patients and healthy blood donors. In this context, Random Forest (RF), J48, AdaBoost, Gradient Boosting Classifiers (GBC), and Light Gradient Boosting Machine (Light GBM) algorithms from bagging and boosting models were used. The most successful classification result was obtained with the Light GBM algorithm as 98.8%, 98.1%, 99.4%, and 0.98%, respectively, in terms of accuracy, precision, recall, and kappa statistics using 10-fold cross-validation.