<![CDATA[Inefficiency in claims submitted by health facilities to BPJS Kesehatan is a problem that needs attention. With the growing number of claims, traditional method such as manual verification is unable to handle enormous amounts of data swiftly. Simultaneously, there is a demand to accelerate the settlement of claims with a limited number of verifiers. One way that can be adopted is to use Machine Learning to detect potential inefficient transactions rapidly. This study compares several Machine Learning algorithms: Random Forest, Gradient Boosting Classifier, Decision Tree, Support Vector Machine, Naive Bayes, CatBoost, and XGBoost. In addition, oversampling and under sampling methods are also used because the dataset is imbalanced. The best results were obtained using the Random Forest + Tomek Links model, which produced an F1 score of 19,53, with the five most influential variables: location of the health facility, participant's age, first-level health facilities diagnosis, participant's primary diagnosis, and health facility type.]]>
Copyrights © 2023
