<![CDATA[This study aims to improve liver disease diagnosis accuracy using the Synthetic Minority Oversampling Technique (SMOTE) and Particle Swarm Optimization (PSO). The dataset used in this study is the Indian Liver Patient Dataset (ILPD), which consists of 583 patient records with 10 input attributes and one class attribute. The main problem in this dataset is the imbalance of class distribution between liver and non-liver patients, which may affect the performance of classification models. The research stages include data preprocessing, missing value imputation, stratified 70:30 train-test splitting, applying SMOTE to the training data, feature selection using PSO, and evaluating several classification models, namely Decision Tree (DT), Support Vector Machine (SVM), Random Forest (RF), Naive Bayes (NB), and Logistic Regression (LR). The experimental results show that SMOTE-RF achieved the best overall performance with an AUC of 0.8610, accuracy of 0.7784, F1-score of 0.7783, and MCC of 0.5570. Meanwhile, SMOTE-PSO-DT improved the Tree model by selecting five important attributes, namely Direct Bilirubin, SGOT, Total Proteins, Albumin, and A/G Ratio. These results indicate that SMOTE effectively improves classification performance, while PSO helps simplify the model through feature selection.]]>
Copyrights © 2026
