<![CDATA[This study addresses the forecasting of visitor numbers to the Pelayanan Statistik Terpadu (PST) unit at the Badan Pusat Statoistic (BPS) Bojonegoro office using a time series approach. Forecasting visitor volume is essential for supporting service capacity planning, resource allocation, and enhancing the operational efficiency of data-driven public services. The study compares three forecasting methods—Seasonal Autoregressive Integrated Moving Average (SARIMA), Holt-Winters Exponential Smoothing, and Prophet—using monthly visitor data from January 2020 to May 2025. The dataset was divided into a training set (January 2020–December 2024) and a test set (January–June 2025). Model evaluation was conducted using R², Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE) metrics. The results indicate that the Prophet model delivered the best performance, achieving an RMSE of 1.44, MAE of 1.34, and MAPE of 0.91—outperforming both Holt-Winters and SARIMA. Prophet's superior performance is attributed to its ability to simultaneously model long-term downward trends and annual seasonal patterns. The findings demonstrate that an adaptive, time series-based forecasting approach can support data-driven decision-making for regional public statistical services.]]>
Copyrights © 2025
