<![CDATA[This study addresses the persistent shortfall between planned and realized budgets for land acquisition financing of Indonesia's National Strategic Projects (Proyek Strategis Nasional/PSN), a shortfall that directly undermines the annual budget absorption rate used as a key performance indicator for the management of state investment funds allocated for land procurement. Method: this study compares two forecasting approaches, the Autoregressive Integrated Moving Average (ARIMA, extended to Seasonal ARIMA where required) and the Artificial Neural Network (ANN), to project the monthly land acquisition financing requirements of PSN across two categories, the Toll Road sector and the Non-Toll Road sector, using monthly realization data managed by the Indonesia Asset Management Agency (Lembaga Manajemen Aset Negara/LMAN) from 2017 through the first quarter of 2026. Forecast accuracy was evaluated using the Mean Absolute Percentage Error (MAPE) and, where the data contained values close to zero, the Symmetric Mean Absolute Percentage Error (sMAPE). Results: the ANN model consistently outperformed its statistical counterpart in both sectors. For the Toll Road sector, ANN achieved a MAPE of 49.01% against 80.93% for ARIMA, while for the Non-Toll Road sector, ANN achieved an sMAPE of 64.08% against 81.53% for SARIMA. Based on the ANN projections for the remaining nine months of 2026, the total financing requirement for land acquisition across all PSN sectors is estimated at Rp10.69 trillion, corresponding to a fund absorption rate of only 40.16% of the currently available budget. Implications: these findings suggest that machine learning approaches offer a more reliable basis for expenditure planning in this context, and that the persistently low absorption rate points to a continuing need for more accurate, data-driven budget execution planning at the ministry and agency level.]]>
Copyrights © 2026
