<![CDATA[Forecasting provincial poverty indicators is important for understanding poverty dynamics across Indonesian provinces. This study aims to forecast the Percentage of Poor Population (P0) at the provincial level for the 2026–2030 period using the Holt method with numerical parameter optimization. Secondary data were obtained from Statistics Indonesia for 2013–2025 and covered five poverty indicators, namely P0, the number of poor people, the poverty line, the Poverty Gap Index (P1), and the Poverty Severity Index (P2). P0 was used as the main forecasting indicator, while the other indicators supported the descriptive analysis. The preprocessing stage included data format standardization, province-name harmonization, selection of Semester I/March observations, removal of national aggregate data, and data completeness checking. Model accuracy was evaluated using MAE and MAPE by comparing the initial Holt model, optimized Holt model, optimized Neural Network, optimized Prophet, and optimized ARIMA. The results show that the optimized Holt model produced the best accuracy, with an MAE of 0.497 and a MAPE of 5.763% on the 2025 test data. The average predicted P0 across provinces decreased from 8.79% in 2026 to 7.56% in 2030. However, forecasts for newly established provinces, particularly in the Papua region, should be interpreted cautiously due to shorter historical time series.]]>
Copyrights © 2026
