Abstract
In Indonesia, the prevalence of chronic kidney disease (CKD) is increasing, especially among individuals with a history of diabetes and hypertension. This makes early detection and timely medical intervention very important to reduce the burden of this disease. Machine learning-based approaches have been widely used in CKD diagnosis systems. The current problem for diagnosis using the Random Forest (RF) method using the Particle Swarm Optimization (PSO) method has not been widely implemented. This study proposes the integration of the RF algorithm with PSO to build a more accurate classification model in detecting CKD. Experiments were conducted using RF and PSO models, evaluated using Accuracy, Precision, Recall, and AUC metrics. The results showed that the RF model without PSO optimization achieved an accuracy of 83.25% and an AUC of 85.00%. After being optimized using PSO, the accuracy increased significantly to 98.00% and the AUC reached 99.00%, indicating a substantial increase in the model's ability to distinguish between CKD sufferers and healthy individuals. These findings strengthen the effectiveness of integrating ensemble learning methods and metaheuristic methods in diagnosing chronic diseases, by providing a basis for developing systems that support more accurate and efficient clinical decision-making.