<![CDATA[The volatility of First-Level Inpatient Care (RITP) claim costs poses a substantial challenge to BPJS Health’s financial management, underscoring the need for accurate forecasting methods. This study employs Artificial Neural Network and Generalized Autoregressive Conditional Heteroscedasticity models to examine volatility dynamics and assess predictive performance. The results indicate that both models capture nonlinear patterns, heteroskedasticity, and temporal dependencies, with evidence that past fluctuations largely influence current volatility. Forecast accuracy is generally high, as reflected in the small discrepancies between predicted and actual values across most provinces. Nevertheless, the models exhibit limitations in capturing extreme peaks and troughs, where abrupt claim variations are not fully represented. These findings highlight the effectiveness of Artificial Neural Networks and Generalized Autoregressive Conditional Heteroscedasticity in modeling claim volatility, while emphasizing the need for model refinement, such as parameter optimization or integration with complementary approaches, to enhance forecasting reliability. ]]>
Copyrights © 2025
