Abstract

Liver disease is one of the health issues of paramount concern in the world today, and early identification is essential to enhance patient outcomes. This paper presents a stacked ensemble learning model that can be used to diagnose liver conditions based on structural medical records having both clinical and biochemical variables. The method combines two strong classifiers namely Random Forest (RF) and AdaBoost into a single ensemble to put into account the interaction among the features and provide greater stability to prediction.In order to enhance generalization, lessen over-fitting and ascertaining confidence assessment, the model was prepared and validated using $\mathbf{k}$-fold cross-validation in a known liver illness benchmark dataset. It is shown that the RF -AdaBoost ensemble with the highest classification accuracy (96.8 percent) outperforms the traditional methods like Decision Trees (91 percent) and Support Vector Machines (88 percent). Other measures such as precision, recall, F1-score and the area under the ROC curve were used to verify that the ensemble was performing better.This suggests that the suggested hybrid model is efficient in dealing with the issue of class imbalance and complex feature interdependence, therefore, can be used as a potential tool in the early diagnosis of liver disease and become an important addition to the clinical decision support systems.

Concepts :

Access to Document

SDGs

Citations by Year