Improving Liver Disease Classification Performance Using Random Forest Algorithm with Grid Search Hyperparameter Optimization

Abstract

Liver disease is one of the serious health problems that often develops without obvious early symptoms, making it difficult to detect early and increasing the risk of more severe complications. The main problem in the classification of liver disease is how to obtain a prediction model that has high accuracy and stability when faced with multivariate and complex medical data. Therefore, this study aims to improve the performance of liver disease classification by optimizing the Random Forest algorithm using the Grid Search hyperparameter optimization technique. The dataset used consisted of 1,700 patient data with eleven attributes that included demographic characteristics, lifestyle, medical risk factors, and liver function test results. The pre-processing stage includes data checking, feature normalization, and data sharing using a ratio of 80% trained data and 20% stratified test data. The Random Forest model is built as a base model and further optimized using Grid Search with a cross-validation scheme to obtain the best combination of hyperparameters. Model performance evaluation was carried out using accuracy, precision, recall, and F1-score metrics. The results showed that the baseline Random Forest model produced an accuracy of 89.71%, while after optimization using Grid Search, the accuracy increased to 91.18% accompanied by an increase in precision, recall, and F1-score values. These improvements show that the hyperparameter optimization process plays an important role in improving the generalization capabilities and stability of the model. The findings of this study confirm that the integration of Random Forest and Grid Search can be an effective approach to the classification of liver diseases and make a significant contribution to the development of medical decision support systems and the application of Machine Learning techniques in the field of health informatics.