<![CDATA[Digital government transformation requires analytical capabilities that not only process data but also generate predictive information to support policy decision-making. This study develops IC-Gov as a Government Decision Intelligence approach that integrates Machine Learning, evidence retrieval, and AI-based synthesis to support the analysis of regional development priorities. The modeling process employs four development indicators: the number of 4G blank spots, the percentage of roads in good condition, the education index, and GRDP per capita, with development priorities classified into low, medium, and high categories. Three algorithms—Random Forest, XGBoost, and Gradient Boosting—were evaluated using accuracy, macro precision, macro recall, macro F1-score, confusion matrix, and ROC-AUC. The results show that Random Forest achieved an accuracy, macro precision, macro recall, and macro F1-score of 1.00, followed by Gradient Boosting with an accuracy of 0.95 and a macro F1-score of 0.95, and XGBoost with an accuracy of 0.87 and a macro F1-score of 0.85. The ROC-AUC values were 1.000, 1.000, and 0.992, respectively. Feature importance analysis indicates that GRDP per capita has the highest relative contribution across all three models. The prediction results are subsequently integrated into IC-Gov as predictive insights, combined with supporting evidence and AI-based synthesis to generate more contextual analytical information. The findings demonstrate that integrating predictive analytics and evidence within a Government Decision Intelligence framework has the potential to strengthen data-driven regional development policy decision support while maintaining human decision-makers as the final authority]]>
